ISL 214E - Statistics II

Prof. Dr. Raziye Selim

selimraziy@itu.edu.tr
Week 1 Part 1 Sampling Distributions
▰ Research questions about management, economics, business,
science (physics, pyschology…)
▰ Statistical tests about them

▰ Knowledge about statistical tests.

▰ Statistical inferences

▰ Parametric tests ( µ, π, σ2 )

▰ Non-parametric Tests ( , relationships between variables)

• Sampling Distributions
Mean E()= µ
• Sampling dist. of the mean St Dev = (St. Error)
Shape ~ Normal

Infinite pop. • Central Limit Theorem

Finite pop. = • If x is not normal
~N if n>30

• Sampling dist. of the proportion ( p )

Mean
St Dev
Shape
• Sampling dist. of the variance

• Confidence Interval
( σ known )
( σ unknown )
 SAMPLING DISTRIBUTIONS

• A major goal of data analysis is to make statistical inferences.

• To use samples to estimate population parameters.

The sample mean

Ex: As the operation manager is interested in using the sample mean calculated from a
sample of cereal boxes for estimating the mean weight contained in a population of boxes.

The sample proportion.

Ex: A poll estimates actual proportion of the votes

• Hypothetically, to use the sample statistic to estimate the population parameter, one should examine every possible
sample that could occur. If this selection of all possible samples were actually done,
the distribution of the results would be referred to as a sampling distribution.
 Sampling Distribution of the Mean
The arithmetic mean is said to be unbiased because the average of all the possible sample means will be equal to the
population mean µ . If we want to show this property emprically:

Example:
Suppose that there are 5 children in an Intellectual Health Center. Their ages are 6, 8, 10, 12, 14.

µ = = 10 f
x f 3

6 1
σ2 = = 8 8 1 2

10 1
12 1 1
14 1
0
6 Uniform
8 Distribution
10 12 14

If samples of two children are selected with replacement from this population; N n = 52 = 25 possible samples can be
selected. These possible sample outcomes are shown in following Table.
 6 8 10 12 14

(6,6) (6,8) (6,10) (6,12) (6,14)

6
6 7 8 9 10
(8,6) (8,8) (8,10) (8,12) (8,14)
8
7 8 9 10 11
(10,6) (10,8) (10,10) (10,12) (10,14)
10
8 9 10 11 12
(12,6) (12,8) (12,10) (12,12) (12,14)
12
9 10 11 12 13
(14,6) (14,8) (14,10) (14,12) (14,14)
14
10 11 12 13 14
 All possible’s and their frequencies;
f p(x)

f(x̅ ) 6 1 0.04
6
7 2 0.08
5
8 3 0.12
4
9 4 0.16
3 10 5 0.20
2 11 4 0.16
1 12 3 0.12
0
13 2 0.08
6 7 8 9 10 11 12 13 14 x̅ 14 1 0.04
25 1.00

The mean of these values:

= = = 10
• The average of all sample means is equal to population mean. As a result, the sample arithmetic mean is an

unbiased estimator of the population mean.

 The variance of the sampling distribution

The variation of the sampling distribution ;

= =4

The sample means are less variable than the population data.

Why?

• A population consists of individual outcomes that can take on a wide range of

values from extremely small to extremely large. However if an extreme value
falls into the sample, the effect will be reduced because it is being averaged
in with all the other values in the sample.

• As the sample size increases, the effect of a single extreme value gets even
smaller because it is being averaged with more observations.
“ The standard deviation of the sampling
distribution is called as standard error.
The variance: =
The standard error: =
 • The standard error of the mean is equal to the standard deviation in the
population σ divided by the square root of the sample size n.

«As the sample size increases, the standard error of the mean decreases.»

What distribution will the sample mean follow?

If sampling is done from a population that is normally distributed with mean µ and standard
deviation σ, the sampling distribution of the mean will also be normally distributed
for any size n with mean = µ and standard error of the mean .

1- ~N

2- =µ

3- =
As a summary:
Sampling Distributions

• is unbiased estimator of µ .
Population Distribution:
Example: Ages N=5 6, 8, 10, 12, 14
2 f

µ = = 10 σ2 = = 8 1

0
6 8 10 12 14
X

With Replace Samples (Infinite Population) If n/N >= 0.05

Without Replace Samples (Finite Population)
f Sampling Distribution:
6 1
6 f(x̅ ) ~N f
7 2 5
7 1
= 10
4
8 3
9 4
3
2
=µ 8 1
= =3
1 9 2
10 5 0
=
11 4
6 7 8 9 10 11 12 13 14
x̅ 10 2 = • =3
11 2
12 3 = 10 Standard Error 12 1
13 2
14 1
=4 = =2 13 1
Example:

• At Oxford Cereal Company plants, thousands of boxes of cereal are filled during each
eight-hour shifts. Boxes are supposed to contain an average 368 grams of cereal. Because
of the nature of the process, there is variation in the cereal weight from box to box.

• Because it is too time-consuming, costly and inefficient to weight every single box, you
must take a sample of boxes and make a decision regarding the likehood that the cereal-
filling process is working properly.

• Each time a sample of boxes is selected and individual boxes are weighted, a sample
mean is calculated. You need to decide the likelihood that such an is could have been
randomly drawn from a population whose true mean µ is 368 grams.
If a sample of 25 boxes is randomly selected without replacement from the thousands of boxes.
Given that the standard deviation of the cereal-filling process is 15 grams.,

The standard error of the mean:

= = =3

If sample of 25 boxes is randomly selected, if the average weight is computed from this sample, what type of
result could be expected? Do you think that the sample mean could be 368 grams? 200 grams? 365 grams?

The sample is a miniature representation of the population. If the values in the population are normally
distributed, the values in the sample should be approximately normally distributed.

If the population mean is 368 grams, the sample mean has a good chance of being close to 368 grams.
How can you determine the probability that the sample of 25 boxes will have a mean below 365 grams?

From the normal distribution: z=

Finding z for the sampling distribution of the mean:

z= = = -1.00

The area is 15.87% of all the possible samples of size 25 have a sample mean below 365 gram.
 A Reminder:
This is not the same as saying that a certain percentage of individual boxes will have
less than 365 grams of cereal.

This is: z = = = = -0.20

42,7% of the individual boxes are expected to contain less than 365 grams.

42,7% of the individual boxes are expected to contain

less than 365 grams.

15
The effect of sample size n on the clustering of means in the sampling distribution;

If n is equal 100

= = 1.5
If a sample of 100 boxes is selected, what’s the likelihood of obtaining a sample mean below 365 grams?

Z= = -2.00
z = -2  2.28 % of the samples of size 100 would be expected to have means below 365 grams.

Sometimes you need to find the interval within which a fixed proportion of the sample’s means fall.
17
 Population
What sample size is large enough? 30

When the sample size is at least 30,

the sampling distribution of the mean
will be approximately normal.

Sampling Distribution

18