What is Statistics

09 March 2023 14:56

Statistics is a branch of mathematics that involves collecting,

analysing, interpreting, and presenting data. It provides tools and
methods to understand and make sense of large amounts of data
and to draw conclusions and make decisions based on the data.

In practice, statistics is used in a wide range of fields, such as

business, economics, social sciences, medicine, and engineering. It is
used to conduct research studies, analyse market trends, evaluate
the effectiveness of treatments and interventions, and make
forecasts and predictions.

Examples:

1. Business - Data Analysis(Identifying customer behavior) and

Demand Forecasting
2. Medical - Identify efficacy of new medicines(Clinical trials),
Identifying risk factor for diseases(Epidemiology)
3. Government & Politics - Conducting surveys, Polling
4. Environmental Science - Climate research

Session 1 on Descriptive Statistics Page 1

Types of Statistics
09 March 2023 14:57

Descriptive statistics deals with the Inferential statistics deals with making
collection, organization, analysis, conclusions and predictions about a population
interpretation, and presentation of data. based on a sample. It involves the use of
It focuses on summarizing and describing probability theory to estimate the likelihood of
the main features of a set of data, without certain events occurring, hypothesis testing to
making inferences or predictions about the determine if a certain claim about a population is
larger population. supported by the data, and regression analysis to
examine the relationships between variables

Session 1 on Descriptive Statistics Page 2

Population Vs Sample
09 March 2023 14:57

Population refers to the entire group of individuals or objects that we are

interested in studying. It is the complete set of observations that we want to make
inferences about. For example, the population might be all the students in a
particular school or all the cars in a particular city.

A sample, on the other hand, is a subset of the population. It is a smaller group of

individuals or objects that we select from the population to study. Samples are
used to estimate characteristics of the population, such as the mean or the
proportion with a certain attribute. For example, we might randomly select 100
students.

Examples

1. All cricket fans vs fans who were present in the stadium

2. All students vs who visit college for lectures

Things to be careful about which creating samples

1. Sample Size
2. Random
3. Representative

Parameter Vs Statistics

A parameter is a characteristic of a population, while a statistic is a characteristic

of a sample. Parameters are generally unknown and are estimated using statistics.
The goal of statistical inference is to use the information obtained from the sample
to make inferences about the population parameters.

Session 1 on Descriptive Statistics Page 3

Inferential Statistics
09 March 2023 14:57

Inferential statistics is a branch of statistics that deals with making inferences or predictions
about a larger population based on a sample of data. It involves using statistical techniques to
test hypotheses and draw conclusions from data. Some of the topics that come under
inferential statistics are:

1. Hypothesis testing: This involves testing a hypothesis about a population parameter based
on a sample of data. For example, testing whether the mean height of a population is
different from a given value.
2. Confidence intervals: This involves estimating the range of values that a population
parameter could take based on a sample of data. For example, estimating the population
mean height within a given confidence level.
3. Analysis of variance (ANOVA): This involves comparing means across multiple groups to
determine if there are any significant differences. For example, comparing the mean
height of individuals from different regions.
4. Regression analysis: This involves modelling the relationship between a dependent
variable and one or more independent variables. For example, predicting the sales of a
product based on advertising expenditure.
5. Chi-square tests: This involves testing the independence or association between two
categorical variables. For example, testing whether gender and occupation are
independent variables.
6. Sampling techniques: This involves ensuring that the sample of data is representative of
the population. For example, using random sampling to select individuals from a
population.
7. Bayesian statistics: This is an alternative approach to statistical inference that involves
updating beliefs about the probability of an event based on new evidence. For example,
updating the probability of a disease given a positive test result.

Why ML is closely associated with statistics?

Session 1 on Descriptive Statistics Page 4

Types of Data
09 March 2023 14:57

Session 1 on Descriptive Statistics Page 5

Measure of Central Tendency
09 March 2023 14:58

A measure of central tendency is a statistical measure that represents a typical or

central value for a dataset. It provides a summary of the data by identifying a
single value that is most representative of the dataset as a whole.

Session 1 on Descriptive Statistics Page 6

1. Mean
09 March 2023 16:35

Mean: The mean is the sum of all values in the dataset divided by the number of
values.

Session 1 on Descriptive Statistics Page 7

2. Median
09 March 2023 16:36

Median: The median is the middle value in the dataset when the data is arranged
in order.

Session 1 on Descriptive Statistics Page 8

3. Mode
09 March 2023 16:36

Mode: The mode is the value that appears most frequently in the dataset.

Session 1 on Descriptive Statistics Page 9

4. Weighted Mean
09 March 2023 16:39

Weighted Mean: The weighted mean is the sum of the products of each value and
its weight, divided by the sum of the weights. It is used to calculate a mean when
the values in the dataset have different importance or frequency.

Session 1 on Descriptive Statistics Page 10

5. Trimmed Mean
10 March 2023 09:37

A trimmed mean is calculated by removing a certain percentage of the

smallest and largest values from the dataset and then taking the mean
of the remaining values. The percentage of values removed is called the
trimming percentage.

Session 1 on Descriptive Statistics Page 11

Measure of Dispersion
09 March 2023 14:58

A measure of dispersion is a statistical measure that describes the spread or

variability of a dataset. It provides information about how the data is distributed
around the central tendency (mean, median or mode) of the dataset.

Session 1 on Descriptive Statistics Page 12

1. Range
09 March 2023 16:36

Range: The range is the difference between the maximum and minimum values in
the dataset. It is a simple measure of dispersion that is easy to calculate but can be
affected by outliers.

Session 1 on Descriptive Statistics Page 13

2. Variance
09 March 2023 16:36

Variance: The variance is the average of the squared differences between each
data point and the mean. It measures the average distance of each data point
from the mean and is useful in comparing the dispersion of datasets with different
means.

X-mean (X-mean)^2
3 3-3 0
2 2-3 1
1 1-3 4
5 5-3 4
4 4-3 1

Mean Absolute Deviation

Session 1 on Descriptive Statistics Page 14

3. Standard Deviation
09 March 2023 16:37

Standard Deviation: The standard deviation is the square root of the variance. It is
a widely used measure of dispersion that is useful in describing the shape of a
distribution.

Session 1 on Descriptive Statistics Page 15

4. Coefficient of Variation
09 March 2023 16:37

Coefficient of Variation (CV): The CV is the ratio of the standard deviation to the
mean expressed as a percentage. It is used to compare the variability of datasets
with different means and is commonly used in fields such as biology, chemistry,
and engineering.

The coefficient of variation (CV) is a statistical measure that expresses the amount
of variability in a dataset relative to the mean. It is a dimensionless quantity that is
expressed as a percentage.

The formula for calculating the coefficient of variation is:

CV = (standard deviation / mean) x 100%

Session 1 on Descriptive Statistics Page 16

Graphs for Univariate Analysis
09 March 2023 14:58

Session 1 on Descriptive Statistics Page 17

1. Categorical - Frequency Distribution Table & Cumulative Frequency
09 March 2023 16:50

A frequency distribution table is a table that summarizes the number of times (or
frequency) that each value occurs in a dataset.

Let's say we have a survey of 200 people and we ask them about their favourite
type of vacation, which could be one of six categories: Beach, City, Adventure,
Nature, Cruise, or Other

Relative frequency is the proportion or percentage of a category in a dataset or

sample. It is calculated by dividing the frequency of a category by the total number
of observations in the dataset or sample.

Cumulative frequency is the running total of frequencies of a variable or category

in a dataset or sample. It is calculated by adding up the frequencies of the current
category and all previous categories in the dataset or sample.

Session 1 on Descriptive Statistics Page 18

2. Numerical - Frequency Distribution Table & Histogram
09 March 2023 16:52

Shapes of Histogram

Session 1 on Descriptive Statistics Page 19

Session 1 on Descriptive Statistics Page 20
Graphs for Bivariate Analysis
09 March 2023 14:59

Session 1 on Descriptive Statistics Page 21

1. Categorical - Categorical
09 March 2023 16:58

Contingency Table/Crosstab

A contingency table, also known as a cross-tabulation or crosstab, is a type of table

used in statistics to summarize the relationship between two categorical variables.
A contingency table displays the frequencies or relative frequencies of the
observed values of the two variables, organized into rows and columns.

Session 1 on Descriptive Statistics Page 22

2. Numerical - Numerical
09 March 2023 16:58

Scatter Plot

Session 1 on Descriptive Statistics Page 23

3. Categorical - Numerical
09 March 2023 16:58

Session 1 on Descriptive Statistics Page 24

Recap
13 March 2023 18:56

Session 2 on Descriptive Statistics Page 1

Quantiles and Percentiles
13 March 2023 06:57

Quantiles are statistical measures used to divide a set of numerical data into
equal-sized groups, with each group containing an equal number of observations.

Quantiles are important measures of variability and can be used to: understand
distribution of data, summarize and compare different datasets. They can also be
used to identify outliers.

There are several types of quantiles used in statistical analysis, including:

a. Quartiles: Divide the data into four equal parts, Q1 (25th percentile), Q2
(50th percentile or median), and Q3 (75th percentile).

b. Deciles: Divide the data into ten equal parts, D1 (10th percentile), D2
(20th percentile), ..., D9 (90th percentile).

c. Percentiles: Divide the data into 100 equal parts, P1 (1st percentile), P2
(2nd percentile), ..., P99 (99th percentile).

d. Quintiles: Divides the data into 5 equal parts

Things to remember while calculating these measures:

1. Data should be sorted from low to high
2. You are basically finding the location of an observation
3. They are not actual values in the data
4. All other tiles can be easily derived from Percentiles

Percentile
A percentile is a statistical measure that represents the percentage of observations in a
dataset that fall below a particular value. For example, the 75th percentile is the value below
which 75% of the observations in the dataset fall.

Formula to calculate the percentile value:

PL =

where:

• PL = the desired percentile value location

• N = the total number of observations in the dataset
• p = the percentile rank (expressed as a percentage)

Example:

Find the 75th percentile score from the below data

78, 82, 84, 88, 91, 93, 94, 96, 98, 99

Step1 - Sort the data

78, 82, 84, 88, 91, 93, 94, 96, 98, 99

Session 2 on Descriptive Statistics Page 2

Percentile of a value
Percentile rank =

X = number of values below the given value

Y = number of values equal to the given value

N = total number of values in the dataset

78, 82, 84, 88, 91, 93, 94, 96, 98, 99

Session 2 on Descriptive Statistics Page 3

5 number summary
13 March 2023 06:57

The five-number summary is a descriptive statistic that provides a summary of a

dataset. It consists of five values that divide the dataset into four equal parts, also
known as quartiles. The five-number summary includes the following values:

1. Minimum value: The smallest value in the dataset.

2. First quartile (Q1): The value that separates the lowest 25% of the data from
the rest of the dataset.

3. Median (Q2): The value that separates the lowest 50% from the highest 50%
of the data.

4. Third quartile (Q3): The value that separates the lowest 75% of the data from
the highest 25% of the data.

5. Maximum value: The largest value in the dataset.

The five-number summary is often represented visually using a box plot, which
displays the range of the dataset, the median, and the quartiles.
The five-number summary is a useful way to quickly summarize the central
tendency, variability, and distribution of a dataset.

Interquartile Range
The interquartile range (IQR) is a measure of variability that is based on the five-number
summary of a dataset. Specifically, the IQR is defined as the difference between the third
quartile (Q3) and the first quartile (Q1) of a dataset.

Session 2 on Descriptive Statistics Page 4

Boxplots
13 March 2023 06:57

1. What is a boxplot
A box plot, also known as a box-and-whisker plot, is a graphical representation of a
dataset that shows the distribution of the data. The box plot displays a summary of the
data, including the minimum and maximum values, the first quartile (Q1), the median
(Q2), and the third quartile (Q3).

2. How to create a boxplot with example

Session 2 on Descriptive Statistics Page 5

1. Benefits of a Boxplot
○ Easy way to see the distribution of data
○ Tells about skewness of data
○ Can identify outliers
○ Compare 2 categories of data

2. Side by side boxplot

Session 2 on Descriptive Statistics Page 6

Session 2 on Descriptive Statistics Page 7
Scatterplots
13 March 2023 06:58

Session 2 on Descriptive Statistics Page 8

Covariance
13 March 2023 06:57

• What problem does Covariance solve?

• What is covariance and how is it interpreted?

Covariance is a statistical measure that describes the degree to which two variables are
linearly related. It measures how much two variables change together, such that when
one variable increases, does the other variable also increase, or does it decrease?

If the covariance between two variables is positive, it means that the variables tend to
move together in the same direction. If the covariance is negative, it means that the
variables tend to move in opposite directions. A covariance of zero indicates that the
variables are not linearly related.

• How is it calculated?

Session 2 on Descriptive Statistics Page 9

Exp(x) Salary(y) X-Xmean Y-Ymean (X-Xmean)*(Y-Ymean)
2 1
5 2
8 5
12 12
13 10

Backlogs(x) package(y) X-Xmean Y-Ymean (X-Xmean)*(Y-Ymean)

2 10
5 12
8 5
12 2
13 1

Session 2 on Descriptive Statistics Page 10

Backlogs(x) package(y) X-Xmean Y-Ymean (X-Xmean)*(Y-Ymean)
2 10
5 10
8 10
12 10
13 10

• Disadvantages of using Covariance

One limitation of covariance is that it does not tell us about the strength of the
relationship between two variables, since the magnitude of covariance is affected by the
scale of the variables.

• Covariance of a variable with itself

Session 2 on Descriptive Statistics Page 11

Session 2 on Descriptive Statistics Page 12
Correlation
13 March 2023 06:58

1. What problem does Correlation solve?

Can we quantify this weak and strong relationship?

2. What is correlation?
Correlation refers to a statistical relationship between two or more variables.
Specifically, it measures the degree to which two variables are related and
how they tend to change together.

Correlation is often measured using a statistical tool called the correlation

coefficient, which ranges from -1 to 1. A correlation coefficient of -1 indicates
a perfect negative correlation, a correlation coefficient of 0 indicates no
correlation, and a correlation coefficient of 1 indicates a perfect positive
correlation.

Session 2 on Descriptive Statistics Page 13

Correlation and Causation
13 March 2023 18:31

The phrase "correlation does not imply causation" means that just because
two variables are associated with each other, it does not necessarily mean that
one causes the other. In other words, a correlation between two variables
does not necessarily imply that one variable is the reason for the other
variable's behaviour.

Suppose there is a positive correlation between the number of firefighters

present at a fire and the amount of damage caused by the fire. One might be
tempted to conclude that the presence of firefighters causes more damage.
However, this correlation could be explained by a third variable - the severity
of the fire. More severe fires might require more firefighters to be present, and
also cause more damage.

Thus, while correlations can provide valuable insights into how different
variables are related, they cannot be used to establish causality. Establishing
causality often requires additional evidence such as experiments, randomized
controlled trials, or well-designed observational studies.

Session 2 on Descriptive Statistics Page 14

Visualizing Multiple Variables
13 March 2023 06:58

1. 3D Scatter Plots

2. Hue Parameter

3. Facetgrids

Session 2 on Descriptive Statistics Page 15

4. Jointplots

5. Pairplots

Session 2 on Descriptive Statistics Page 16

6. Bubble Plots

Session 2 on Descriptive Statistics Page 17

Random Variables
15 March 2023 11:43

• What are Algebraic Variables?

In Algebra a variable, like x, is an unknown value

• What are Random Variables in Stats and Probability?

A Random Variable is a set of possible values from a random experiment.

• Types of Random Variables?

Session 3 - Descriptive Statistics Page 1

Probability Distributions
15 March 2023 11:53

1. What are Probability Distributions?

A probability distribution is a list of all of the possible outcomes of a random variable
along with their corresponding probability values.

Problem with Distribution?

In many scenarios, the number of outcomes can be much larger and hence a table would
be tedious to write down. Worse still, the number of possible outcomes could be infinite,
in which case, good luck writing a table for that.

Example - Height of people, Rolling 10 dice together

Solution - Function?

What if we use a mathematical function to model the relationship between outcome and
probability?

Note - A lot of time Probability Distribution and Probability Distribution Functions are

Session 3 - Descriptive Statistics Page 2

 Note - A lot of time Probability Distribution and Probability Distribution Functions are
used interchangeably.

1. Types of Probability Distributions

Famous Probability Distributions

Session 3 - Descriptive Statistics Page 3

Why are Probability Distributions important?
- Gives an idea about the shape/distribution of the data.
- And if our data follows a famous distribution then we automatically know a lot about the
data.

A note on Parameters
Parameters in probability distributions are numerical values that determine the shape,
location, and scale of the distribution.

Different probability distributions have different sets of parameters that determine their
shape and characteristics, and understanding these parameters is essential in statistical
analysis and inference.

Session 3 - Descriptive Statistics Page 4

Probability Distribution Functions
15 March 2023 20:08

A probability distribution function (PDF) is a mathematical function that describes

the probability of obtaining different values of a random variable in a particular
probability distribution.

Session 3 - Descriptive Statistics Page 5

Probability Mass Function (PMF)
15 March 2023 15:25

PMF stands for Probability Mass Function. It is a mathematical function that

describes the probability distribution of a discrete random variable.

The PMF of a discrete random variable assigns a probability to each possible value
of the random variable. The probabilities assigned by the PMF must satisfy two
conditions:

a. The probability assigned to each value must be non-negative (i.e., greater

than or equal to zero).
b. The sum of the probabilities assigned to all possible values must equal 1.

Examples

https://en.wikipedia.org/wiki/Bernoulli_distribution

https://en.wikipedia.org/wiki/Binomial_distribution

Session 3 - Descriptive Statistics Page 6

Cumulative Distribution Function(CDF) of PMF
15 March 2023 20:09

The cumulative distribution function (CDF) F(x) describes the probability that a
random variable X with a given probability distribution will be found at a value less
than or equal to x

F(x) = P(X <= x)

Examples:

https://en.wikipedia.org/wiki/Bernoulli_distribution

https://en.wikipedia.org/wiki/Binomial_distribution

Session 3 - Descriptive Statistics Page 7

Probability Density Function (PDF)
15 March 2023 15:25

PDF stands for Probability Density Function. It is a mathematical function that

describes the probability distribution of a continuous random variable.

1. Why Probability Density and why not Probability?

2. What does the area of this graph represents?
3. How to calculate Probability then?
4. Examples of PDF
a. https://en.wikipedia.org/wiki/Normal_distribution
b. https://en.wikipedia.org/wiki/Log-normal_distribution
c. https://en.wikipedia.org/wiki/Poisson_distribution
5. How is graph calculated?

Session 3 - Descriptive Statistics Page 8

Density Estimation
16 March 2023 06:54

Density estimation is a statistical technique used to estimate the probability

density function (PDF) of a random variable based on a set of observations or data.
In simpler terms, it involves estimating the underlying distribution of a set of data
points.

Density estimation can be used for a variety of purposes, such as hypothesis

testing, data analysis, and data visualization. It is particularly useful in areas such
as machine learning, where it is often used to estimate the probability distribution
of input data or to model the likelihood of certain events or outcomes.

There are various methods for density estimation, including parametric and non-
parametric approaches. Parametric methods assume that the data follows a
specific probability distribution (such as a normal distribution), while non-
parametric methods do not make any assumptions about the distribution and
instead estimate it directly from the data.

Commonly used techniques for density estimation include kernel density

estimation (KDE), histogram estimation, and Gaussian mixture models (GMMs).
The choice of method depends on the specific characteristics of the data and the
intended use of the density estimate.

Session 3 - Descriptive Statistics Page 9

Parametric Density Estimation
16 March 2023 06:54

Parametric density estimation is a method of estimating the probability density

function (PDF) of a random variable by assuming that the underlying distribution
belongs to a specific parametric family of probability distributions, such as the
normal, exponential, or Poisson distributions.

Session 3 - Descriptive Statistics Page 10

Non-Parametric Density Estimation (KDE)
16 March 2023 06:55

But sometimes the distribution is not clear or it's not one of the famous distributions.

Non-parametric density estimation is a statistical technique used to estimate the probability

density function of a random variable without making any assumptions about the underlying
distribution. It is also referred to as non-parametric density estimation because it does not
require the use of a predefined probability distribution function, as opposed to parametric
methods such as the Gaussian distribution.

The non-parametric density estimation technique involves constructing an estimate of the

probability density function using the available data. This is typically done by creating a kernel
density estimate

Non-parametric density estimation has several advantages over parametric density

estimation. One of the main advantages is that it does not require the assumption of a
specific distribution, which allows for more flexible and accurate estimation in situations
where the underlying distribution is unknown or complex. However, non-parametric density
estimation can be computationally intensive and may require more data to achieve accurate
estimates compared to parametric methods.

Session 3 - Descriptive Statistics Page 11

Kernel Density Estimate(KDE)
16 March 2023 16:08

The KDE technique involves using a kernel function to smooth out the data and create a
continuous estimate of the underlying density function.

Session 3 - Descriptive Statistics Page 12

Cumulative Distribution Function(CDF) of PDF
15 March 2023 15:25

Session 3 - Descriptive Statistics Page 13

How to use PDF and CDF in Data Analysis
15 March 2023 20:10

Session 3 - Descriptive Statistics Page 14

2D Probability Density Plots
16 March 2023 06:50

Session 3 - Descriptive Statistics Page 15

Recap
21 March 2023 14:50

Session on Normal Distri Page 1

How to use PDF in Data Science
20 March 2023 18:11

Session on Normal Distri Page 2

2D Density Plots
20 March 2023 18:11

Session on Normal Distri Page 3

Normal Distribution
20 March 2023 18:06

1. What is normal distribution?

Normal distribution, also known as Gaussian distribution, is a probability distribution that
is commonly used in statistical analysis. It is a continuous probability distribution that is
symmetrical around the mean, with a bell-shaped curve.

-> Tail
-> Asymptotic in nature
-> Lots of points near the mean and very few far away

The normal distribution is characterized by two parameters: the mean (μ) and the
standard deviation (σ). The mean represents the centre of the distribution, while the
standard deviation represents the spread of the distribution.

Denoted as:

Why is it so important?

Commonality in Nature: Many natural phenomena follow a normal distribution, such as

the heights of people, the weights of objects, the IQ scores of a population, and many
more. Thus, the normal distribution provides a convenient way to model and analyse such
data.

PDF Equation of Normal Distribution

Parameters in Normal Distribution

https://samp-suman-normal-dist-visualize-app-lkntug.streamlit.app/

Equation in detail:

Session on Normal Distri Page 4

Session on Normal Distri Page 5
Standard Normal Variate
20 March 2023 18:08

• What is Standard Normal Variate

A Standard Normal Variate(Z) is a standardized form of the normal distribution with mean
= 0 and standard deviation = 1.

Standardizing a normal distribution allows us to compare different distributions with each

other, and to calculate probabilities using standardized tables or software.

Equation:

• How to transform a normal distribution to Standard Normal Variate

Refer Python code

Kya Fayda Standardize karne ka?

Suppose the heights of adult males in a certain population follow a normal distribution
with a mean of 68 inches and a standard deviation of 3 inches. What is the probability
that a randomly selected adult male from this population is taller than 72 inches?

• What are Z-tables

A z-table tells you the area underneath a normal distribution curve, to the left of the z-
score
https://www.ztable.net/

For a Normal Distribution X~(u,std) what percent of population lie between mean and 1
standard deviation, 2 std and 3 std?

Session on Normal Distri Page 6

standard deviation, 2 std and 3 std?

Session on Normal Distri Page 7

Properties of Normal Distribution
20 March 2023 18:06

1. Symmetricity
The normal distribution is symmetric about its mean, which means that the probability of
observing a value above the mean is the same as the probability of observing a value below
the mean. The bell-shaped curve of the normal distribution reflects this symmetry.

2. Measures of Central Tendencies are equal

3. Empirical Rule
The normal distribution has a well-known empirical rule, also called the 68-95-99.7 rule,
which states that approximately 68% of the data falls within one standard deviation of the
mean, about 95% of the data falls within two standard deviations of the mean, and about
99.7% of the data falls within three standard deviations of the mean.

4. The area under the curve

Session on Normal Distri Page 8

Session on Normal Distri Page 9
Skewness
20 March 2023 18:07

• What is skewness?
A normal distribution is a bell-shaped, symmetrical distribution with a specific
mathematical formula that describes how the data is spread out. Skewness indicates that
the data is not symmetrical, which means it is not normally distributed.

Skewness is a measure of the asymmetry of a probability distribution. It is a statistical

measure that describes the degree to which a dataset deviates from the normal
distribution.

In a symmetrical distribution, the mean, median, and mode are all equal. In contrast, in a
skewed distribution, the mean, median, and mode are not equal, and the distribution
tends to have a longer tail on one side than the other.

Skewness can be positive, negative, or zero. A positive skewness means that the tail of
the distribution is longer on the right side, while a negative skewness means that the tail
is longer on the left side. A zero skewness indicates a perfectly symmetrical distribution.

The greater the skew the greater the distance between mode, median and mode.

• How skewness is calculated?

• Python Example

• Interpretation

Session on Normal Distri Page 10

CDF of Normal Distribution
20 March 2023 18:07

Session on Normal Distri Page 11

Use in Data Science
20 March 2023 18:08

• Outlier detection
• Assumptions on data for ML algorithms -> Linear Regression and GMM
• Hypothesis Testing
• Central Limit Theorem

Session on Normal Distri Page 12

Recap
23 March 2023 18:27

Session on Non-Gaussian Distribution Page 1

Kurtosis
23 March 2023 13:18

• What is Kurtosis?
Kurtosis is the 4th statistical moment. In probability theory and statistics, kurtosis
(meaning "curved, arching") is a measure of the "tailedness" of the probability
distribution of a real-valued random variable. Like skewness, kurtosis describes a
particular aspect of a probability distribution.

• False notation about Kurtosis

https://en.wikipedia.org/wiki/Kurtosis

• Formula

• Practical Use-case

In finance, kurtosis risk refers to the risk associated with the possibility of extreme
outcomes or "fat tails" in the distribution of returns of a particular asset or portfolio.

If a distribution has high kurtosis, it means that there is a higher likelihood of

extreme events occurring, either positive or negative, compared to a normal
distribution.

In finance, kurtosis risk is important to consider because it indicates that there is a

greater probability of large losses or gains occurring, which can have significant
implications for investors. As a result, investors may want to adjust their investment
strategies to account for kurtosis risk.

• Excess Kurtosis & Types

Excess kurtosis is a measure of how much more peaked or flat a distribution is

Session on Non-Gaussian Distribution Page 2

• Excess Kurtosis & Types
Excess kurtosis is a measure of how much more peaked or flat a distribution is
compared to a normal distribution, which is considered to have a kurtosis of 0. It is
calculated by subtracting 3 from the sample kurtosis coefficient.

Types of Kurtosis

Leptokurtic

A distribution with positive excess kurtosis is called leptokurtic. "Lepto-" means

"slender". In terms of shape, a leptokurtic distribution has fatter tails. This indicates
that there are more extreme values or outliers in the distribution.

Example - Assets with positive excess kurtosis are riskier and more volatile than those
with a normal distribution, and they may experience sudden price movements that
can result in significant gains or losses.

Platykurtic

A distribution with negative excess kurtosis is called platykurtic. "Platy-" means

"broad". In terms of shape, a platykurtic distribution has thinner tails. This indicates
that there are fewer extreme values or outliers in the distribution.

Assets with negative excess kurtosis are less risky and less volatile than those with a
normal distribution, and they may experience more gradual price movements that
are less likely to result in large gains or losses.

Mesokurtic

Distributions with zero excess kurtosis are called mesokurtic. The most prominent
example of a mesokurtic distribution is the normal distribution family, regardless of
the values of its parameters.

Mesokurtic is a term used to describe a distribution with a excess kurtosis of 0,

indicating that it has the same degree of "peakedness" or "flatness" as a normal
distribution.

Example -
In finance, a mesokurtic distribution is considered to be the ideal distribution for
assets or portfolios, as it represents a balance between risk and return.

Session on Non-Gaussian Distribution Page 3

QQ Plot
23 March 2023 13:19

• How to find if a given distribution is normal or not?

○ Visual inspection: One of the easiest ways to check for normality is to visually inspect
a histogram or a density plot of the data. A normal distribution has a bell-shaped
curve, which means that the majority of the data falls in the middle, and the tails
taper off symmetrically. If the distribution looks approximately bell-shaped, it is likely
to be normal.

○ QQ Plot: Another way to check for normality is to create a normal probability plot
(also known as a Q-Q plot) of the data. A normal probability plot plots the observed
data against the expected values of a normal distribution. If the data points fall along
a straight line, the distribution is likely to be normal.

○ Statistical tests: There are several statistical tests that can be used to test for
normality, such as the Shapiro-Wilk test, the Anderson-Darling test, and the
Kolmogorov-Smirnov test. These tests compare the observed data to the expected
values of a normal distribution and provide a p-value that indicates whether the data
is likely to be normal or not. A p-value less than the significance level (usually 0.05)
suggests that the data is not normal.

• What is a QQ Plot and how is it plotted?

A QQ plot (quantile-quantile plot) is a graphical tool used to assess the similarity of
the distribution of two sets of data. It is particularly useful for determining whether a
set of data follows a normal distribution.

In a QQ plot, the quantiles of the two sets of data are plotted against each other. The
quantiles of one set of data are plotted on the x-axis, while the quantiles of the other
set of data are plotted on the y-axis. If the two sets of data have the same
distribution, the points on the QQ plot will fall on a straight line. If the two sets of
data do not have the same distribution, the points will deviate from the straight line.

• Python example
• https://www.statsmodels.org/dev/generated/statsmodels.graphics.gofplots.qqplot.html

• How to interpret QQ plots

Session on Non-Gaussian Distribution Page 4

1.

1.

Session on Non-Gaussian Distribution Page 5

 1.

• Does QQ plot only detect normal distribution?

Session on Non-Gaussian Distribution Page 6

Uniform Distribution
23 March 2023 13:19

• What is Uniform Distribution and it's types

In probability theory and statistics, a uniform distribution is a probability distribution

where all outcomes are equally likely within a given range. This means that if you were to
select a random value from this range, any value would be as likely as any other value.

Types

Denoted as

• Examples
a. The height of a person randomly selected from a group of individuals whose heights
range from 5'6" to 6'0" would follow a continuous uniform distribution.
b. The time it takes for a machine to produce a product, where the production time
ranges from 5 to 10 minutes, would follow a continuous uniform distribution.
c. The distance that a randomly selected car travels on a tank of gas, where the
distance ranges from 300 to 400 miles, would follow a continuous uniform
distribution.
d. The weight of a randomly selected apple from a basket of apples that weighs
between 100 and 200 grams, would follow a continuous uniform distribution.

• PDF CDF and Graphs

https://en.wikipedia.org/wiki/Continuous_uniform_distribution

• Skewness

• Application in Machine learning and Data Science

a. Random initialization: In many machine learning algorithms, such as neural networks
and k-means clustering, the initial values of the parameters can have a significant
impact on the final result. Uniform distribution is often used to randomly initialize
the parameters, as it ensures that all values in the range have an equal probability of
being selected.

Session on Non-Gaussian Distribution Page 7

a. Random initialization: In many machine learning algorithms, such as neural networks
and k-means clustering, the initial values of the parameters can have a significant
impact on the final result. Uniform distribution is often used to randomly initialize
the parameters, as it ensures that all values in the range have an equal probability of
being selected.

b. Sampling: Uniform distribution can also be used for sampling. For example, if you
have a dataset with an equal number of samples from each class, you can use
uniform distribution to randomly select a subset of the data that is representative of
all the classes.

c. Data augmentation: In some cases, you may want to artificially increase the size of
your dataset by generating new examples that are similar to the original data.
Uniform distribution can be used to generate new data points that are within a
specified range of the original data.

d. Hyperparameter tuning: Uniform distribution can also be used in hyperparameter

tuning, where you need to search for the best combination of hyperparameters for a
machine learning model. By defining a uniform prior distribution for each
hyperparameter, you can sample from the distribution to explore the
hyperparameter space.

Session on Non-Gaussian Distribution Page 8

Log Normal Distribution
23 March 2023 13:19

In probability theory and statistics, a lognormal distribution is a heavy tailed continuous

probability distribution of a random variable whose logarithm is normally distributed.

Examples

• The length of comments posted in Internet discussion forums follows a log-normal

distribution.
• Users' dwell time on online articles (jokes, news etc.) follows a log-normal distribution.
• The length of chess games tends to follow a log-normal distribution.
• In economics, there is evidence that the income of 97%–99% of the population is
distributed log-normally.

Denoted as

PDF Equation

CDF

Skewness

How to check if a random variable is log normally distributed?

Session on Non-Gaussian Distribution Page 9

Skewness

How to check if a random variable is log normally distributed?

Session on Non-Gaussian Distribution Page 10

Pareto Distribution
23 March 2023 13:19

Pareto Distribution

The Pareto distribution is a type of probability distribution that is commonly used to model the
distribution of wealth, income, and other quantities that exhibit a similar power-law behaviour

What is Power Law

In mathematics, a power law is a functional relationship between two variables, where one
variable is proportional to a power of the other. Specifically, if y and x are two variables
related by a power law, then the relationship can be written as:

y = k * x^a

Vilfredo Pareto originally used this distribution to describe the allocation of wealth among
individuals since it seemed to show rather well the way that a larger portion of the wealth of
any society is owned by a smaller percentage of the people in that society. He also used it to
describe distribution of income. This idea is sometimes expressed more simply as the Pareto
principle or the "80-20 rule" which says that 20% of the population controls 80% of the wealth

Graph & Parameters

Examples

• The sizes of human settlements (few cities, many hamlets/villages)

• File size distribution of Internet traffic which uses the TCP protocol (many smaller files,
few larger ones)

CDF

Session on Non-Gaussian Distribution Page 11

Skewness

How to detect if a distribution is Pareto Distribution?

Session on Non-Gaussian Distribution Page 12

Transformations
23 March 2023 13:24

Session on Non-Gaussian Distribution Page 13

Some Terms
30 March 2023 07:09

Population Vs Sample
Population: A population is the entire group or set of individuals, objects, or events that a
researcher wants to study or draw conclusions about. It can be people, animals, plants, or
even inanimate objects, depending on the context of the study. The population usually
represents the complete set of possible data points or observations.

Sample: A sample is a subset of the population that is selected for study. It is a smaller group
that is intended to be representative of the larger population. Researchers collect data from
the sample and use it to make inferences about the population as a whole. Since it is often
impractical or impossible to collect data from every member of a population, samples are used
as an efficient and cost-effective way to gather information.

Parameter Vs Estimate
Parameter: A parameter is a numerical value that describes a characteristic of a population.
Parameters are usually denoted using Greek letters, such as μ (mu) for the population mean or
σ (sigma) for the population standard deviation. Since it is often difficult or impossible to
obtain data from an entire population, parameters are usually unknown and must be
estimated based on available sample data.

Sta s c sta s c is a numerical value that describes a characteris c of a sample, hich is a

subset of the popula on. y using sta s cs calculated from a representa ve sample,
researchers can make inferences about the unkno n respec ve parameter of the popula on.
ommon sta s cs include the sample mean (denoted by , pronounced bar ), the sample
median, and the sample standard devia on (denoted by s).

Inferential Statistics
Inferential statistics is a branch of statistics that focuses on making predictions, estimations, or
generalizations about a larger population based on a sample of data taken from that
population. It involves the use of probability theory to make inferences and draw conclusions
about the characteristics of a population by analysing a smaller subset or sample.

The key idea behind inferential statistics is that it is often impractical or impossible to collect
data from every member of a population, so instead, we use a representative sample to make
inferences about the entire group. Inferential statistical techniques include hypothesis testing,
confidence intervals, and regression analysis, among others.

These methods help researchers answer questions like:

a. Is there a significant difference between two groups?

b. Can we predict the outcome of a variable based on the values of other variables?
c. What is the relationship between two or more variables?

Inferential statistics are widely used in various fields, such as economics, social sciences,
medicine, and natural sciences, to make informed decisions and guide policy based on limited
data.

Session on Confidence Interval Page 1

Point Estimate
30 March 2023 07:19

A point estimate is a single value, calculated from a sample, that serves as the best guess or
approximation for an unknown population parameter, such as the mean or standard
deviation. Point estimates are often used in statistics when we want to make inferences about
a population based on a sample.

Session on Confidence Interval Page 2

Confidence Interval
30 March 2023 07:18

Confidence interval, in simple words, is a range of values within which we expect a particular
population parameter, like a mean, to fall. It's a way to express the uncertainty around an
estimate obtained from a sample of data.

Confidence level, usually expressed as a percentage like 95%, indicates how sure we are that
the true value lies within the interval.

Confidence Interval = Point Estimate Margin of Error

Ways to calculate CI:

Confidence Interval is created for Parameters and not statistics. Statistics help us get the
confidence interval for a parameter.

Examples of CT usage

Session on Confidence Interval Page 3

Session on Confidence Interval Page 4
Confidence Interval (Sigma Known)
30 March 2023 07:13

Assumptions

1. Random sampling: The data must be collected using a random sampling method to
ensure that the sample is representative of the population. This helps to minimize biases
and ensures that the results can be generalized to the entire population.

2. Known population standard deviation The population standard deviation (σ) must be
known or accurately estimated. In practice, the population standard deviation is often
unknown, and the sample standard deviation (s) is used as an estimate. However, if the
sample size is large enough, the sample standard deviation can provide a reasonably
accurate approximation.

3. Normal distribution or large sample size: The Z-procedure assumes that the underlying
population is normally distributed. However, if the population distribution is not normal,
the Central Limit Theorem can be applied when the sample size is large (usually, sample
size n ≥ 30 is considered large enough). According to the Central Limit Theorem, the
sampling distribution of the sample mean will approach a normal distribution as the
sample size increases, regardless of the shape of the population distribution.

A (1 - alpha)*100% Confidence Interval for mu:

Session on Confidence Interval Page 5

Session on Confidence Interval Page 6
Session on Confidence Interval Page 7
Interpreting Confidence Interval
30 March 2023 08:33

A confidence interval is a range of values within which a population parameter, such as the
population mean, is estimated to lie with a certain level of confidence. The confidence interval
provides an indication of the precision and uncertainty associated with the estimate. To
interpret the confidence interval values, consider the following points:

1. Confidence level: The confidence level (commonly set at 90%, 95%, or 99%) represents
the probability that the confidence interval will contain the true population parameter if
the sampling and estimation process were repeated multiple times. For example, a 95%
confidence interval means that if you were to draw 100 different samples from the
population and calculate the confidence interval for each, approximately 95 of those
intervals would contain the true population parameter.

2. Interval range: The width of the confidence interval gives an indication of the precision of
the estimate. A narrower confidence interval suggests a more precise estimate of the
population parameter, while a wider interval indicates greater uncertainty. The width of
the interval depends on the sample size, variability in the data, and the desired level of
confidence.

3. Interpretation: To interpret the confidence interval values, you can say that you are "X%
confident that the true population parameter lies within the range (lower limit, upper
limit)." Keep in mind that this statement is about the interval, not the specific point
estimate, and it refers to the confidence level you chose when constructing the interval.
What is the trade-off

Session on Confidence Interval Page 8

Factors Affecting Margin of Error
30 March 2023 07:15

1. Confidence Level (1-alpha)

2. Sample Size
3. Population Standard Deviation

Session on Confidence Interval Page 9

Confidence Interval (Sigma not known)
30 March 2023 07:15

Using the t procedure

Assumptions

1. Random sampling: The data must be collected using a random sampling method to
ensure that the sample is representative of the population. This helps to minimize biases
and ensures that the results can be generalized to the entire population.

2. Sample standard deviation The population standard deviation (σ) is unkno n, and the
sample standard deviation (s) is used as an estimate. The t-distribution is specifically
designed to account for the additional uncertainty introduced by using the sample
standard deviation instead of the population standard deviation.

3. Approximately normal distribution: The t-procedure assumes that the underlying

population is approximately normally distributed, or the sample size is large enough for
the Central Limit Theorem to apply. If the population distribution is heavily skewed or has
extreme outliers, the t-procedure may not be accurate, and non-parametric methods
should be considered.

4. Independent observations: The observations in the sample should be independent of

each other. In other words, the value of one observation should not influence the value of
another observation. This is particularly important when working with time series data or
data with inherent dependencies.

Session on Confidence Interval Page 10

Session on Confidence Interval Page 11
Session on Confidence Interval Page 12
Student's T Distribution
30 March 2023 07:16

Student's t-distribution, or simply the t-distribution, is a probability distribution that arises when
estimating the mean of a normally distributed population when the sample size is small and the
population standard deviation is unknown. It was introduced by William Sealy Gosset, who
published under the pseudonym "Student."

The t-distribution is similar to the normal distribution (also known as the Gaussian distribution or
the bell curve) but has heavier tails. The shape of the t-distribution is determined by the degrees of
freedom, which is closely related to the sample size (degrees of freedom = sample size - 1). As the
degrees of freedom increase (i.e., as the sample size increases), the t-distribution approaches the
normal distribution.

In hypothesis testing and confidence interval estimation, the t-distribution is used in place of the
normal distribution when the sample size is small (usually less than 30) and the population standard
deviation is unknown. The t-distribution accounts for the additional uncertainty that arises from
estimating the population standard deviation using the sample standard deviation.

To use the t-distribution in practice, you look up critical t-values from a t-distribution table, which
provides values corresponding to specific degrees of freedom and confidence levels (e.g., 95%
confidence). These critical t-values are then used to calculate confidence intervals or perform
hypothesis tests.

Session on Confidence Interval Page 13

Titanic Case Study
31 March 2023 18:00

Session on Confidence Interval Page 14

Bernoulli Distribution
27 March 2023 16:06

Bernoulli distribution is a probability distribution that models a binary outcome, where the
outcome can be either success (represented by the value 1) or failure (represented by the
value 0). The Bernoulli distribution is named after the Swiss mathematician Jacob Bernoulli,
who first introduced it in the late 1600s.

The Bernoulli distribution is characterized by a single parameter, which is the probability of

success, denoted by p. The probability mass function (PMF) of the Bernoulli distribution is:

The Bernoulli distribution is commonly used in machine learning for modelling

binary outcomes, such as whether a customer will make a purchase or not,
whether an email is spam or not, or whether a patient will have a certain disease
or not.

Session on Central Limit Theorem Page 1

Binomial Distribution
27 March 2023 16:36

Binomial distribution is a probability distribution that describes the number of

successes in a fixed number of independent Bernoulli trials with two possible
outcomes (often called "success" and "failure"), where the probability of success
is constant for each trial. The binomial distribution is characterized by two
parameters: the number of trials n and the probability of success p.

The Probability of anyone watching this lecture in the future and then liking it is 0.5. What is the
probability that:

1. No-one out of 3 people will like it

1. 1 out of 3 people will like it

1. 2 out of 3 people will like it

1. 3 out of 3 people will like it

PDF Formula:

Graph of PDF:

Criteria:

Session on Central Limit Theorem Page 2

Criteria:

1. The process consists of n trials

2. Only 2 exclusive outcomes are possible, a success and a failure.
3. P(success) = p and P(failure) = 1-p and it is fixed from trial to trial
4. The trials are independent.

1. Binary classification problems: In binary classification problems, we often model

the probability of an event happening as a binomial distribution. For example, in a
spam detection system, we may model the probability of an email being spam or
not spam using a binomial distribution.

2. Hypothesis testing: In statistical hypothesis testing, we use the binomial

distribution to calculate the probability of observing a certain number of
successes in a given number of trials, assuming a null hypothesis is true. This can
be used to make decisions about whether a certain hypothesis is supported by
the data or not.

3. Logistic regression: Logistic regression is a popular machine learning algorithm

used for classification problems. It models the probability of an event happening
as a logistic function of the input variables. Since the logistic function can be
viewed as a transformation of a linear combination of inputs, the output of
logistic regression can be thought of as a binomial distribution.

4. A/B testing: A/B testing is a common technique used to compare two different
versions of a product, web page, or marketing campaign. In A/B testing, we
randomly assign individuals to one of two groups and compare the outcomes of
interest between the groups. Since the outcomes are often binary (e.g., click-
through rate or conversion rate), the binomial distribution can be used to model
the distribution of outcomes and test for differences between the groups.

Session on Central Limit Theorem Page 3

Sampling Distribution
27 March 2023 17:10

Sampling distribution is a probability distribution that describes the statistical properties of a

sample statistic (such as the sample mean or sample proportion) computed from multiple
independent samples of the same size from a population.

Why Sampling Distribution is important?

Sampling distribution is important in statistics and machine learning because it allows us to

estimate the variability of a sample statistic, which is useful for making inferences about the
population. By analysing the properties of the sampling distribution, we can compute
confidence intervals, perform hypothesis tests, and make predictions about the population
based on the sample data.

Session on Central Limit Theorem Page 4

Central Limit Theorem
27 March 2023 17:10

The Central Limit Theorem (CLT) states that the distribution of the sample means of a large
number of independent and identically distributed random variables will approach a normal
distribution, regardless of the underlying distribution of the variables.

The conditions required for the CLT to hold are:

1. The sample size is large enough, typically greater than or equal to 30.
2. The sample is drawn from a finite population or an infinite population with a finite
variance.
3. The random variables in the sample are independent and identically distributed.

The CLT is important in statistics and machine learning because it allows us to

make probabilistic inferences about a population based on a sample of data. For
example, we can use the CLT to construct confidence intervals, perform
hypothesis tests, and make predictions about the population mean based on the
sample data. The CLT also provides a theoretical justification for many commonly
used statistical techniques, such as t-tests, ANOVA, and linear regression.

Session on Central Limit Theorem Page 5

Case Study 1 - Titanic Fare
28 March 2023 17:19

Session on Central Limit Theorem Page 6

Case Study - What is the average income of Indians
28 March 2023 15:49

Step-by-step process:

1. Collect multiple random samples of salaries from a representative group

of Indians. Each sample should be large enough (usually, n > 30) to ensure
the CLT holds. Make sure the samples are representative and unbiased to
avoid skewed results.

2. Calculate the sample mean (average salary) and sample standard

deviation for each sample.

3. Calculate the average of the sample means. This value will be your best
estimate of the population mean (average salary of all Indians).

4. Calculate the standard error of the sample means, which is the standard
deviation of the sample means divided by the square root of the number
of samples.

5. Calculate the confidence interval around the average of the sample means
to get a range within which the true population mean likely falls. For a
95% confidence interval:

lower_limit = average_sample_means - 1.96 * standard_error

upper_limit = average_sample_means + 1.96 * standard_error

6. Report the estimated average salary and the confidence interval.

Python code

Remember that the validity of your results depends on the quality of your
data and the representativeness of your samples. To obtain accurate
results, it's crucial to ensure that your samples are unbiased and
representative.

Session on Central Limit Theorem Page 7

Hypothesis Testing
04 April 2023 07:04

A statistical hypothesis test is a method of statistical inference used to decide

whether the data at hand sufficiently support a particular hypothesis. Hypothesis
testing allows us to make probabilistic statements about population parameters.

Session 1 on Hypothesis Testing Page 1

Null and Alternate Hypothesis
04 April 2023 07:09

1. Null hypothesis (H0):

In simple terms, the null hypothesis is a statement that assumes there is no significant
effect or relationship between the variables being studied. It serves as the starting point
for hypothesis testing and represents the status quo or the assumption of no effect until
proven otherwise. The purpose of hypothesis testing is to gather evidence (data) to either
reject or fail to reject the null hypothesis in favour of the alternative hypothesis, which
claims there is a significant effect or relationship.

2. Alternative hypothesis (H1 or Ha):

The alternative hypothesis, is a statement that contradicts the null hypothesis and claims
there is a significant effect or relationship between the variables being studied. It
represents the research hypothesis or the claim that the researcher wants to support
through statistical analysis.

Important Points

• How to decide what will be Null hypothesis and what will be Alternate
Hypothesis(Typically the Null hypothesis says nothing new is happening)

• We try to gather evidence to reject the null hypothesis

• It's important to note that failing to reject the null hypothesis doesn't necessarily mean
that the null hypothesis is true; it just means that there isn't enough evidence to support
the alternative hypothesis.

Hypothesis tests are similar to jury trials, in a sense. In a jury trial, H0 is similar to the not-guilty verdict,
and Ha is the guilty verdict. You assume in a jury trial that the defendant isn’t guilty unless the
prosecution can show beyond a reasonable doubt that he or she is guilty. If the jury says the evidence is
beyond a reasonable doubt, they reject H0, not guilty, in favour of Ha , guilty.

Session 1 on Hypothesis Testing Page 2

Steps involved in Hypothesis Testing
04 April 2023 13:25

Rejection Region Approach

1. Formulate a Null and Alternate hypothesis

2. Select a significance level(This is the probability of rejecting the null hypothesis when it is
actually true, usually set at 0.05 or 0.01)
3. Check assumptions (example distribution)
4. Decide which test is appropriate(Z-test, T-test, Chi-square test, ANOVA)
5. State the relevant test statistic
6. Conduct the test
7. Reject or not reject the Null Hypothesis.
8. Interpret the result

Session 1 on Hypothesis Testing Page 3

Performing a Z test Example 1
04 April 2023 07:15

Suppose a company is evaluating the impact of a new training program on the productivity of
its employees. The company has data on the average productivity of its employees before
implementing the training program. The average productivity was 50 units per day with a
known population standard deviation of 5 units. After implementing the training program, the
company measures the productivity of a random sample of 30 employees. The sample has an
average productivity of 53 units per day. The company wants to know if the new training
program has significantly increased productivity.

Session 1 on Hypothesis Testing Page 4

Example 2
04 April 2023 16:06

Suppose a snack food company claims that their Lays wafer packets contain an
average weight of 50 grams per packet. To verify this claim, a consumer watchdog
organization decides to test a random sample of Lays wafer packets. The
organization wants to determine whether the actual average weight differs
significantly from the claimed 50 grams. The organization collects a random
sample of 40 Lays wafer packets and measures their weights. They find that the
sample has an average weight of 49 grams, with a known population standard
deviation of 4 grams.

Session 1 on Hypothesis Testing Page 5

Rejection Region
04 April 2023 16:21

Significance level - denoted as α (alpha), is a predetermined threshold used in hypothesis

testing to determine whether the null hypothesis should be rejected or not. It represents the
probability of rejecting the null hypothesis when it is actually true, also known as Type 1 error.

The critical region is the region of values that corresponds to the rejection of the null
hypothesis at some chosen probability level.

Problem with Rejection Region Approach

Session 1 on Hypothesis Testing Page 6

Type 1 vs Type 2 Error
04 April 2023 13:29

In hypothesis testing, there are two types of errors that can

occur when making a decision about the null hypothesis: Type
I error and Type II error.

Type-I (False Positive) error occurs when the sample results,

lead to the rejection of the null hypothesis when it is in fact
true.

In other words, it's the mistake of finding a significant effect or

relationship when there is none. The probability of committing
a Type I error is denoted by α (alpha), which is also known as
the significance level. By choosing a significance level,
researchers can control the risk of making a Type I error.

Type-II (False Negative) error occurs when based on the

sample results, the null hypothesis is not rejected when it is in
fact false.

This means that the researcher fails to detect a significant

effect or relationship when one actually exists. The probability
of committing a Type II error is denoted by β (beta).

Trade-off between Type 1 and Type 2 errors

Session 1 on Hypothesis Testing Page 7

One sided vs two sided test
04 April 2023 13:29

One-sided (one-tailed) test: A one-sided test is used when the researcher is interested in
testing the effect in a specific direction (either greater than or less than the value specified in
the null hypothesis). The alternative hypothesis in a one-sided test contains an inequality
(either ">" or "<").

Example: A researcher wants to test whether a new medication increases the average
recovery rate compared to the existing medication.

Two-sided (two-tailed) test: A two-sided test is used when the researcher is interested in
testing the effect in both directions (i.e., whether the value specified in the null hypothesis is
different, either greater or lesser). The alternative hypothesis in a two-sided test contains a
"not equal to" sign (≠).

Example: A researcher wants to test whether a new medication has a different average
recovery rate compared to the existing medication.

The main difference between them lies in the directionality of the alternative hypothesis and
how the significance level is distributed in the critical regions.

Advantages and Disadvantages?

Two-tailed test (two-sided):

Advantages:

1. Detects effects in both directions: Two-tailed tests can detect effects in both directions,
which makes them suitable for situations where the direction of the effect is uncertain or
when researchers want to test for any difference between the groups or variables.

2. More conservative: Two-tailed tests are more conservative because the significance level
(α) is split between both tails of the distribution. This reduces the risk of Type I errors in
cases where the direction of the effect is uncertain.

Disadvantages:

1. Less powerful: Two-tailed tests are generally less powerful than one-tailed tests because
the significance level (α) is divided between both tails of the distribution. This means the
test requires a larger effect size to reject the null hypothesis, which could lead to a higher
risk of Type II errors (failing to reject the null hypothesis when it is false).

2. Not appropriate for directional hypotheses: Two-tailed tests are not ideal for cases where
the research question or hypothesis is directional, as they test for differences in both
directions, which may not be of interest or relevance.

One-tailed test (one-sided):

Advantages:

1. More powerful: One-tailed tests are generally more powerful than two-tailed tests, as the
entire significance level (α) is allocated to one tail of the distribution. This means that the
test is more likely to detect an effect in the specified direction, assuming the effect exists.

2. Directional hypothesis: One-tailed tests are appropriate when there is a strong theoretical
or practical reason to test for an effect in a specific direction.

Disadvantages:

1. Missed effects: One-tailed tests can miss effects in the opposite direction of the specified
alternative hypothesis. If an effect exists in the opposite direction, the test will not be
able to detect it, which could lead to incorrect conclusions.

2. Increased risk of Type I error: One-tailed tests can be more prone to Type I errors if the
effect is actually in the opposite direction than the one specified in the alternative
hypothesis.

Session 1 on Hypothesis Testing Page 8

Session 1 on Hypothesis Testing Page 9
Where can be Hypothesis Testing Applied?
04 April 2023 07:15

1. Testing the effectiveness of interventions or treatments: Hypothesis testing can be used to

determine whether a new drug, therapy, or educational intervention has a significant effect
compared to a control group or an existing treatment.

2. Comparing means or proportions: Hypothesis testing can be used to compare means or

proportions between two or more groups to determine if there's a significant difference.
This can be applied to compare average customer satisfaction scores, conversion rates, or
employee performance across different groups.

3. Analysing relationships between variables: Hypothesis testing can be used to evaluate the
association between variables, such as the correlation between age and income or the
relationship between advertising spend and sales.

4. Evaluating the goodness of fit: Hypothesis testing can help assess if a particular theoretical
distribution (e.g., normal, binomial, or Poisson) is a good fit for the observed data.

5. Testing the independence of categorical variables: Hypothesis testing can be used to

determine if two categorical variables are independent or if there's a significant association
between them. For example, it can be used to test if there's a relationship between the
type of product and the likelihood of it being returned by a customer.

6. A/B testing: In marketing, product development, and website design, hypothesis testing is
often used to compare the performance of two different versions (A and B) to determine
which one is more effective in terms of conversion rates, user engagement, or other
metrics.

Session 1 on Hypothesis Testing Page 10

Hypothesis Testing ML Applications
04 April 2023 16:50

1. Model comparison: Hypothesis testing can be used to compare the performance of

different machine learning models or algorithms on a given dataset. For example, you can
use a paired t-test to compare the accuracy or error rate of two models on multiple cross-
validation folds to determine if one model performs significantly better than the other.

2. Feature selection: Hypothesis testing can help identify which features are significantly
related to the target variable or contribute meaningfully to the model's performance. For
example, you can use a t-test, chi-square test, or ANOVA to test the relationship between
individual features and the target variable. Features with significant relationships can be
selected for building the model, while non-significant features may be excluded.

3. Hyperparameter tuning: Hypothesis testing can be used to evaluate the performance of a

model trained with different hyperparameter settings. By comparing the performance of
models with different hyperparameters, you can determine if one set of hyperparameters
leads to significantly better performance.

4. Assessing model assumptions: In some cases, machine learning models rely on certain
statistical assumptions, such as linearity or normality of residuals in linear regression.
Hypothesis testing can help assess whether these assumptions are met, allowing you to
determine if the model is appropriate for the data.

Session 1 on Hypothesis Testing Page 11

Recap
06 April 2023 19:43

Session 2 on Hypothesis Testing Page 1

P-value
06 April 2023 06:48

P-value is the probability of getting a sample as or more extreme(having more evidence

against H0) than our own sample given the Null Hypothesis(H0) is true.

In simple words p-value is a measure of the strength of the evidence against the Null
Hypothesis that is provided by our sample data.

Session 2 on Hypothesis Testing Page 2

Interpreting p-value
06 April 2023 08:25

With significance value

Without significance value

1. Very small p-values (e.g., p < 0.01) indicate strong evidence against the null hypothesis,
suggesting that the observed effect or difference is unlikely to have occurred by chance
alone.
2. Small p-values (e.g., 0.01 ≤ p < 0.05) indicate moderate evidence against the null
hypothesis, suggesting that the observed effect or difference is less likely to have
occurred by chance alone.
3. Large p-values (e.g., 0.05 ≤ p < 0.1) indicate weak evidence against the null hypothesis,
suggesting that the observed effect or difference might have occurred by chance alone,
but there is still some level of uncertainty.
4. Very large p-values (e.g., p ≥ 0.1) indicate weak or no evidence against the null
hypothesis, suggesting that the observed effect or difference is likely to have occurred by
chance alone.

Session 2 on Hypothesis Testing Page 3

P-value in context of Z-test
06 April 2023 07:08

Suppose a company is evaluating the impact of a new training program on the productivity of its employees. The
company has data on the average productivity of its employees before implementing the training program. The
average productivity was 50 units per day. After implementing the training program, the company measures the
productivity of a random sample of 30 employees. The sample has an average productivity of 53 units per day and
the pop std is 4. The company wants to know if the new training program has significantly increased productivity.

Suppose a snack food company claims that their Lays wafer packets contain an average weight of 50 grams per
packet. To verify this claim, a consumer watchdog organization decides to test a random sample of Lays wafer
packets. The organization wants to determine whether the actual average weight differs significantly from the
claimed 50 grams. The organization collects a random sample of 40 Lays wafer packets and measures their
weights. They find that the sample has an average weight of 49 grams, with a pop standard deviation of 5
grams.

Session 2 on Hypothesis Testing Page 4

Session 2 on Hypothesis Testing Page 5
T-tests
06 April 2023 14:14

A t-test is a statistical test used in hypothesis testing to compare the means of two samples or
to compare a sample mean to a known population mean. The t-test is based on the t-
distribution, which is used when the population standard deviation is unknown and the
sample size is small.

There are three main types of t-tests:

One-sample t-test: The one-sample t-test is used to compare the mean of a single sample to a
known population mean. The null hypothesis states that there is no significant difference
between the sample mean and the population mean, while the alternative hypothesis states
that there is a significant difference.

Independent two-sample t-test: The independent two-sample t-test is used to compare the
means of two independent samples. The null hypothesis states that there is no significant
difference between the means of the two samples, while the alternative hypothesis states that
there is a significant difference.

Paired t-test (dependent two-sample t-test): The paired t-test is used to compare the means of
two samples that are dependent or paired, such as pre-test and post-test scores for the same
group of subjects or measurements taken on the same subjects under two different
conditions. The null hypothesis states that there is no significant difference between the
means of the paired differences, while the alternative hypothesis states that there is a
significant difference.

Session 2 on Hypothesis Testing Page 6

Single Sample t-test
06 April 2023 14:14

A one-sample t-test checks whether a sample mean differs from the population mean.

Assumptions for a single sample t-test

1. Normality - Population from which the sample is drawn is normally distributed

2. Independence - The observations in the sample must be independent, which means that
the value of one observation should not influence the value of another observation.
3. Random Sampling - The sample must be a random and representative subset of the
population.
4. Unknown population std - The population std is not known.

Suppose a manufacturer claims that the average weight of their new chocolate bars is 50
grams, we highly doubt that and want to check this so we drew out a sample of 25 chocolate
bars and measured their weight, the sample mean came out to be 49.7 grams and the sample
std deviation was 1.2 grams. Consider the significance level to be 0.05

Session 2 on Hypothesis Testing Page 7

Python Case Study 1
06 April 2023 17:27

Session 2 on Hypothesis Testing Page 8

Independent 2 sample t-test
06 April 2023 14:15

An independent two-sample t-test, also known as an unpaired t-test, is a statistical method

used to compare the means of two independent groups to determine if there is a significant
difference between them.

Assumptions for the test:

1. Independence of observations: The two samples must be independent, meaning there is

no relationship between the observations in one group and the observations in the other
group. The subjects in the two groups should be selected randomly and independently.

2. Normality: The data in each of the two groups should be approximately normally
distributed. The t-test is considered robust to mild violations of normality, especially
when the sample sizes are large (typically n ≥ 30) and the sample sizes of the two groups
are similar. If the data is highly skewed or has substantial outliers, consider using a non-
parametric test, such as the Mann-Whitney U test.

3. Equal variances (Homoscedasticity): The variances of the two populations should be

approximately equal. This assumption can be checked using F-test for equality of
variances. If this assumption is not met, you can use Welch's t-test, which does not
require equal variances.

4. Random sampling: The data should be collected using a random sampling method from
the respective populations. This ensures that the sample is representative of the
population and reduces the risk of selection bias.

Suppose a website owner claims that there is no difference in the average time spent on their
website between desktop and mobile users. To test this claim, we collect data from 30
desktop users and 30 mobile users regarding the time spent on the website in minutes. The
sample statistics are as follows:

desktop users = [12, 15, 18, 16, 20, 17, 14, 22, 19, 21, 23, 18, 25, 17, 16, 24, 20, 19, 22, 18, 15,
14, 23, 16, 12, 21, 19, 17, 20, 14]

mobile_users = [10, 12, 14, 13, 16, 15, 11, 17, 14, 16, 18, 14, 20, 15, 14, 19, 16, 15, 17, 14, 12,
11, 18, 15, 10, 16, 15, 13, 16, 11]

Desktop users:
○ Sample size (n1): 30
○ Sample mean (mean1): 18.5 minutes
○ Sample standard deviation (std_dev1): 3.5 minutes

Mobile users:
○ Sample size (n2): 30
○ Sample mean (mean2): 14.3 minutes
○ Sample standard deviation (std_dev2): 2.7 minutes

We will use a significance level (α) of 0.05 for the hypothesis test.

Session 2 on Hypothesis Testing Page 9

Session 2 on Hypothesis Testing Page 10
Python Case Study 2
06 April 2023 17:27

Session 2 on Hypothesis Testing Page 11

Paired 2 sample t-test
06 April 2023 14:21

A paired two-sample t-test, also known as a dependent or paired-samples t-test, is a statistical

test used to compare the means of two related or dependent groups.

Common scenarios where a paired two-sample t-test is used include:

1. Before-and-after studies: Comparing the performance of a group before and after an

intervention or treatment.

2. Matched or correlated groups: Comparing the performance of two groups that are
matched or correlated in some way, such as siblings or pairs of individuals with similar
characteristics.

Assumptions

1. Paired observations: The two sets of observations must be related or paired in some way,
such as before-and-after measurements on the same subjects or observations from
matched or correlated groups.

2. Normality: The differences between the paired observations should be approximately

normally distributed. This assumption can be checked using graphical methods (e.g.,
histograms, Q-Q plots) or statistical tests for normality (e.g., Shapiro-Wilk test). Note that
the t-test is generally robust to moderate violations of this assumption when the sample
size is large.

3. Independence of pairs: Each pair of observations should be independent of other pairs. In

other words, the outcome of one pair should not affect the outcome of another pair. This
assumption is generally satisfied by appropriate study design and random sampling.

Let's assume that a fitness center is evaluating the effectiveness of a new 8 -week weight loss
program. They enroll 15 participants in the program and measure their weights before and
after the program. The goal is to test whether the new weight loss program leads to a
significant reduction in the participants' weight.

Before the program:

[80, 92, 75, 68, 85, 78, 73, 90, 70, 88, 76, 84, 82, 77, 91]
After the program:
[78, 93, 81, 67, 88, 76, 74, 91, 69, 88, 77, 81, 80, 79, 88]

Significance level (α) = 0.05

Session 2 on Hypothesis Testing Page 12

Session 2 on Hypothesis Testing Page 13
Chi Square Distribution
07 April 2023 10:29

The Chi-Square distribution, also written as χ² distribution, is a continuous

probability distribution that is widely used in statistical hypothesis testing,
particularly in the context of goodness-of-fit tests and tests for independence in
contingency tables. It arises when the sum of the squares of independent
standard normal random variables follows this distribution.

The Chi-Square distribution has a single parameter, the degrees of freedom (df),
which influences the shape and spread of the distribution. The degrees of
freedom are typically associated with the number of independent variables or
constraints in a statistical problem.

Some key properties of the Chi-Square distribution are:

a. It is a continuous distribution, defined for non-negative values.

b. It is positively skewed, with the degree of skewness decreasing as the
degrees of freedom increase.
c. The mean of the Chi-Square distribution is equal to its degrees of
freedom, and its variance is equal to twice the degrees of freedom.
d. As the degrees of freedom increase, the Chi-Square distribution
approaches the normal distribution in shape.

The Chi-Square distribution is used in various statistical tests, such as the Chi-
Square goodness-of-fit test, which evaluates whether an observed frequency
distribution fits an expected theoretical distribution, and the Chi-Square test for
independence, which checks the association between categorical variables in a
contingency table.

Session 3 on Hypothesis Testing Page 1

Chi Square Test
07 April 2023 15:03

The Chi-Square test is a statistical hypothesis test used to determine if there is a significant
association between categorical variables or if an observed distribution of categorical data
differs from an expected theoretical distribution. It is based on the Chi-Square (χ²) distribution,
and it is commonly applied in two main scenarios:

1. Chi-Square Goodness-of-Fit Test: This test is used to determine if the observed

distribution of a single categorical variable matches an expected theoretical distribution.
It is often applied to check if the data follows a specific probability distribution, such as
the uniform or binomial distribution.

2. Chi-Square Test for Independence (Chi-Square Test for Association): This test is used to
determine whether there is a significant association between two categorical variables in
a sample.

Session 3 on Hypothesis Testing Page 2

Goodness of Fit Test
07 April 2023 10:29

The Chi-Square Goodness-of-Fit test is a statistical hypothesis test used to

determine if the observed distribution of a single categorical variable
matches an expected theoretical distribution. It helps to evaluate
whether the data follows a specific probability distribution, such as
uniform, binomial, or Poisson distribution, among others. This test is
particularly useful when you want to assess if the sample data is
consistent with an assumed distribution or if there are significant
deviations from the expected pattern.

Steps

The Chi-Square Goodness-of-Fit test involves the following steps:

• Define the null hypothesis (H0) and the alternative hypothesis (H1):
○ H0: The observed data follows the expected theoretical
distribution.
○ H1: The observed data does not follow the expected theoretical
distribution.

• Calculate the expected frequencies for each category based on the

theoretical distribution and the sample size.

• Compute the Chi-Square test statistic (χ²) by comparing the observed

and expected frequencies. The test statistic is calculated as:

• where Oi is the observed frequency in category i, Ei is the expected

frequency in category i, and the summation is taken over all
categories.

• Determine the degrees of freedom (df), which is typically the

number of categories minus one (df = k - 1), where k is the number
of categories.

• Calculate the p-value for the test statistic using the Chi-Square
distribution with the calculated degrees of freedom.

• Compare the test statistic to the critical value or the p-value

Assumptions

1. Independence: The observations in the sample must be independent

of each other. This means that the outcome of one observation
should not influence the outcome of another observation.

2. Categorical data: The variable being analysed must be categorical,

not continuous or ordinal. The data should be divided into mutually
exclusive and exhaustive categories.

3. Expected frequency: Each category should have an expected

frequency of at least 5. This guideline helps ensure that the Chi-
Square distribution is a reasonable approximation for the
distribution of the test statistic. Having small expected frequencies
can lead to an inaccurate estimation of the Chi-Square distribution,
potentially increasing the likelihood of a Type I error (incorrectly
rejecting the null hypothesis) or a Type II error (incorrectly failing to
reject the null hypothesis).

Session 3 on Hypothesis Testing Page 3

 potentially increasing the likelihood of a Type I error (incorrectly
rejecting the null hypothesis) or a Type II error (incorrectly failing to
reject the null hypothesis).

4. Fixed distribution: The theoretical distribution being compared to

the observed data should be specified before the test is conducted.
It is essential to avoid choosing a distribution based on the observed
data, as doing so can lead to biased results.

The Chi-Square Goodness-of-Fit test is a non-parametric test. Non-

parametric tests do not assume that the data comes from a specific
probability distribution or make any assumptions about population
parameters like the mean or standard deviation.

In the Chi-Square Goodness-of-Fit test, we compare the observed

frequencies of the categorical data to the expected frequencies based on
a hypothesized distribution. The test doesn't rely on any assumptions
about the underlying distribution's parameters. Instead, it focuses on
comparing observed counts to expected counts, making it a non-
parametric test.

Session 3 on Hypothesis Testing Page 4

Example 1
07 April 2023 16:26

Suppose we have a six-sided fair die, and we want to test if the die is indeed fair. We roll the
die 60 times and record the number of times each side comes up. We'll use the Chi-Square
Goodness-of-Fit test to determine if the observed frequencies are consistent with a fair die
(i.e., a uniform distribution of the sides).

Observed frequencies:
○ Side 1: 12 times
○ Side 2: 8 times
○ Side 3: 11 times
○ Side 4: 9 times
○ Side 5: 10 times
○ Side 6: 10 times

Session 3 on Hypothesis Testing Page 5

Example 2
07 April 2023 15:26

Suppose a marketing team at a retail company wants to understand the distribution of visits to
their website by day of the week. They have a hypothesis that visits are uniformly distributed
across all days of the week, meaning they expect an equal number of visits on each day. They
collected data on website visits for four weeks and want to test if the observed distribution
matches the expected uniform distribution.

Observed frequencies (number of website visits per day of the week for four weeks):

• Monday: 420
• Tuesday: 380
• Wednesday: 410
• Thursday: 400
• Friday: 410
• Saturday: 430
• Sunday: 390

Session 3 on Hypothesis Testing Page 6

Example 3
07 April 2023 15:27

A survey of 800 families in a village with 4 children each revealed the

following distribution:

Is this data consistent with the result that male and female births are
equally probable?

Session 3 on Hypothesis Testing Page 7

Python Case Study
07 April 2023 15:27

Session 3 on Hypothesis Testing Page 8

Test for Independence
07 April 2023 10:29

The Chi-Square test for independence, also known as the Chi-Square test for association, is a
statistical test used to determine whether there is a significant association between two
categorical variables in a sample. It helps to identify if the occurrence of one variable is
dependent on the occurrence of the other variable, or if they are independent of each other.

The test is based on comparing the observed frequencies in a contingency table (a table that
displays the frequency distribution of the variables) with the frequencies that would be
expected under the assumption of independence between the two variables.

Steps

1. State the null hypothesis (H0) and alternative hypothesis (H1):

○ H0: There is no association between the two categorical variables (they are
independent).
○ H1: There is an association between the two categorical variables (they are
dependent).

2. Create a contingency table with the observed frequencies for each combination of the
categories of the two variables.

3. Calculate the expected frequencies for each cell in the contingency table assuming that
the null hypothesis is true (i.e., the variables are independent).

4. Compute the Chi-Square test statistic:

χ² = Σ [(O_ij - E_ij)² / E_ij]

where O_ij is the observed frequency in each cell and E_ij is the expected frequency.

5. Determine the degrees of freedom: df = (number of rows - 1) * (number of columns - 1)

6. Obtain the critical value or p-value using the Chi-Square distribution table or a statistical
software/calculator with the given degrees of freedom and significance level (commonly
α = 0.05).

7. Compare the test statistic to the critical value or the p-value to the significance level to
decide whether to reject or fail to reject the null hypothesis. If the test statistic is greater
than the critical value, or if the p-value is less than the significance level, we reject the null
hypothesis and conclude that there is a significant association between the two variables.

Assumptions

1. Independence of observations: The observations in the sample should be independent of

each other. This means that the occurrence of one observation should not affect the
occurrence of another observation. In practice, this usually implies that the data should
be collected using a simple random sampling method.

2. Categorical variables: Both variables being tested must be categorical, either ordinal or
nominal. The Chi-Square test for independence is not appropriate for continuous
variables.

3. Adequate sample size: The sample size should be large enough to ensure that the
expected frequency for each cell in the contingency table is sufficient. A common rule of
thumb is that the expected frequency for each cell should be at least 5. If some cells have
expected frequencies less than 5, the test may not be valid, and other methods like
Fisher's exact test may be more appropriate.

4. Fixed marginal totals: The marginal totals (the row and column sums of the contingency
table) should be fixed before the data is collected. This is because the Chi-Square test for
independence assesses the association between the two variables under the assumption
that the marginal totals are fixed and not influenced by the relationship between the
variables.

Session 3 on Hypothesis Testing Page 9

Example 1
07 April 2023 17:50

A researcher wants to investigate if there is an association between the level of education

(categorical variable) and the preference for a particular type of exercise (categorical variable)
among a group of 150 individuals. The researcher collects data and creates the following
contingency table

Session 3 on Hypothesis Testing Page 10

Python Case Study
07 April 2023 15:06

Session 3 on Hypothesis Testing Page 11

Applications in Machine Learning
07 April 2023 17:30

1. Feature selection: Chi-Square test can be used as a filter-based feature selection method to
rank and select the most relevant categorical features in a dataset. By measuring the
association between each categorical feature and the target variable, you can eliminate
irrelevant or redundant features, which can help improve the performance and efficiency
of machine learning models.

2. Evaluation of classification models: For multi-class classification problems, the Chi-Square

test can be used to compare the observed and expected class frequencies in the confusion
matrix. This can help assess the goodness of fit of the classification model, indicating how
well the model's predictions align with the actual class distributions.

3. Analysing relationships between categorical features: In exploratory data analysis, the Chi-
Square test for independence can be applied to identify relationships between pairs of
categorical features. Understanding these relationships can help inform feature
engineering and provide insights into the underlying structure of the data.

4. Discretization of continuous variables: When converting continuous variables into

categorical variables (binning), the Chi-Square test can be used to determine the optimal
number of bins or intervals that best represent the relationship between the continuous
variable and the target variable.

5. Variable selection in decision trees: Some decision tree algorithms, such as the CHAID (Chi-
squared Automatic Interaction Detection) algorithm, use the Chi-Square test to determine
the most significant splitting variables at each node in the tree. This helps construct more
effective and interpretable decision trees.

Session 3 on Hypothesis Testing Page 12

08 April 2023 20:05

Session 4 on Hypothesis Testing Page 1

 F - Distribution
08 April 2023 13:04

1. Continuous probability distribution: The F-distribution is a continuous

probability distribution used in statistical hypothesis testing and
analysis of variance (ANOVA).
2. Fisher-Snedecor distribution: It is also known as the Fisher-Snedecor
distribution, named after Ronald Fisher and George Snedecor, two
prominent statisticians.
3. Degrees of freedom: The F-distribution is defined by two parameters -
the degrees of freedom for the numerator (df1) and the degrees of
freedom for the denominator (df2).
4. Positively skewed and bounded: The shape of the F-distribution is
positively skewed, with its left bound at zero. The distribution's shape
depends on the values of the degrees of freedom.
5. Testing equality of variances: The F-distribution is commonly used to
test hypotheses about the equality of two variances in different
samples or populations.
6. Comparing statistical models: The F-distribution is also used to compare
the fit of different statistical models, particularly in the context of
ANOVA.
7. F-statistic: The F-statistic is calculated by dividing the ratio of two
sample variances or mean squares from an ANOVA table. This value is
then compared to critical values from the F-distribution to determine
statistical significance.
8. Applications: The F-distribution is widely used in various fields of
research, including psychology, education, economics, and the natural
and social sciences, for hypothesis testing and model comparison.

Session 4 on Hypothesis Testing Page 2

One way ANOVA test
08 April 2023 13:12

One-way ANOVA (Analysis of Variance) is a statistical method used to compare the means of
three or more independent groups to determine if there are any significant differences
between them. It is an extension of the t-test, which is used for comparing the means of two
independent groups. The term "one-way" refers to the fact that there is only one independent
variable (factor) with multiple levels (groups) in this analysis.

The primary purpose of one-way ANOVA is to test the null hypothesis that all the group means
are equal. The alternative hypothesis is that at least one group mean is significantly different
from the others.

Steps

• Define the null and alternative hypotheses.

• Calculate the overall mean (grand mean) of all the groups combined and mean of all the
groups individually.
• Calculate the "between-group" and "within-group" sum of squares (SS).
• Find the between group and within group degree of freedoms
• Calculate the "between-group" and "within-group" mean squares (MS) by dividing their
respective sum of squares by their degrees of freedom.
• Calculate the F-statistic by dividing the "between-group" mean square by the "within-
group" mean square.

• Calculate the p-value associated with the calculated F-statistic using the F-distribution and
the appropriate degrees of freedom. The p-value represents the probability of obtaining
an F-statistic as extreme or more extreme than the calculated value, assuming the null
hypothesis is true.
• Choose a significance level (alpha), typically 0.05.
• Compare the calculated p-value with the chosen significance level (alpha).

a. If the p-value is less than or equal to alpha, reject the null hypothesis in favour of the alternative
hypothesis, concluding that there is a significant difference between at least one pair of group
means.
b. If the p-value is greater than alpha, fail to reject the null hypothesis, concluding that there is not
enough evidence to suggest a significant difference between the group means.

It's important to note that one-way ANOVA only determines if there is a significant difference
between the group means; it does not identify which specific groups have significant
differences. To determine which pairs of groups are significantly different, post-hoc tests, such
as Tukey's HSD or Bonferroni, are conducted after a significant ANOVA result.

Session 4 on Hypothesis Testing Page 3

Session 4 on Hypothesis Testing Page 4
Example 1
08 April 2023 13:22

Session 4 on Hypothesis Testing Page 5

Session 4 on Hypothesis Testing Page 6
Geometric Intuition
08 April 2023 13:26

Session 4 on Hypothesis Testing Page 7

Session 4 on Hypothesis Testing Page 8
Assumptions
08 April 2023 16:48

Assumptions

1. Independence: The observations within and between groups should be independent of

each other. This means that the outcome of one observation should not influence the
outcome of another. Independence is typically achieved through random sampling or
random assignment of subjects to groups.

2. Normality: The data within each group should be approximately normally distributed.
While one-way ANOVA is considered to be robust to moderate violations of normality,
severe deviations may affect the accuracy of the test results. If normality is in doubt, non-
parametric alternatives like the Shapiro-wilk test can be considered.

3. Homogeneity of variances: The variances of the populations from which the samples are
drawn should be equal, or at least approximately so. This assumption is known as
homoscedasticity. If the variances are substantially different, the accuracy of the test
results may be compromised. Levene's test or Bartlett's test can be used to assess the
homogeneity of variances. If this assumption is violated, alternative tests such as Welch's
ANOVA can be used.

Session 4 on Hypothesis Testing Page 9

Python Case Study
08 April 2023 13:23

Session 4 on Hypothesis Testing Page 10

Post-hoc Test
08 April 2023 13:23

Post hoc tests, also known as post hoc pairwise comparisons or multiple comparison tests, are
used in the context of ANOVA when the overall test indicates a significant difference among
the group means. These tests are performed after the initial one-way ANOVA to determine
which specific groups or pairs of groups have significantly different means.

The main purpose of post hoc tests is to control the family-wise error rate (FWER) and adjust
the significance level for multiple comparisons to avoid inflated Type I errors. There are
several post hoc tests available, each with different characteristics and assumptions. Some
common post hoc tests include:

1. Bonferroni correction: This method adjusts the significance level (α) by dividing it by the
number of comparisons being made. It is a conservative method that can be applied when
making multiple comparisons, but it may have lower statistical power when a large
number of comparisons are involved.
2. Tukey's HSD (Honestly Significant Difference) test: This test controls the FWER and is used
when the sample sizes are equal and the variances are assumed to be equal across the
groups. It is one of the most commonly used post hoc tests.

When performing post hoc tests, it is essential to choose a test that aligns with the
assumptions of your data (e.g., equal variances, equal sample sizes) and provides an
appropriate balance between controlling Type I errors and maintaining statistical power.

Session 4 on Hypothesis Testing Page 11

Session 4 on Hypothesis Testing Page 12
Why t-test is not used for more than 3 categories?
08 April 2023 13:23

1. Increased Type I error: When you perform multiple comparisons using individual t-tests,
the probability of making a Type I error (false positive) increases. The more tests you
perform, the higher the chance that you will incorrectly reject the null hypothesis in at least
one of the tests, even if the null hypothesis is true for all groups.

2. Difficulty in interpreting results: When comparing multiple groups using multiple t-tests,
the interpretation of the results can become complicated. For example, if you have 4
groups and you perform 6 pairwise t-tests, it can be challenging to interpret and summarize
the overall pattern of differences among the groups.

3. Inefficiency: Using multiple t-tests is less efficient than using a single test that accounts for
all groups, such as one-way ANOVA. One-way ANOVA uses the information from all the
groups simultaneously to estimate the variability within and between the groups, which
can lead to more accurate conclusions.

Session 4 on Hypothesis Testing Page 13

Applications in Machine Learning
08 April 2023 13:27

1. Hyperparameter tuning: When selecting the best hyperparameters for a machine learning
model, one-way ANOVA can be used to compare the performance of models with different
hyperparameter settings. By treating each hyperparameter setting as a group, you can
perform one-way ANOVA to determine if there are any significant differences in
performance across the various settings.

2. Feature selection: One-way ANOVA can be used as a univariate feature selection method to
identify features that are significantly associated with the target variable, especially when
the target variable is categorical with more than two levels. In this context, the one-way
ANOVA is performed for each feature, and features with low p-values are considered to be
more relevant for prediction.

3. Algorithm comparison: When comparing the performance of different machine learning

algorithms, one-way ANOVA can be used to determine if there are any significant
differences in their performance metrics (e.g., accuracy, F1 score, etc.) across multiple runs
or cross-validation folds. This can help you decide which algorithm is the most suitable for a
specific problem.

4. Model stability assessment: One-way ANOVA can be used to assess the stability of a
machine learning model by comparing its performance across different random seeds or
initializations. If the model's performance varies significantly between different
initializations, it may indicate that the model is unstable or highly sensitive to the choice of
initial conditions.

Session 4 on Hypothesis Testing Page 14