CHAPTER 11:

RESEARCH
ANALYSIS;
QUANTITATIVE
DATA
THE GOOD
RESEARCH GUIDE BY
MARTYN DENSCOMBE

Slides dev: Musicha NS

⦿ Sources of quantitative data • Types of
quantitative data • Preparing quantitative
data for analysis • Grouping the data •
Describing the mid-point or average •
Describing the spread of data • Looking for
patterns and relationships in the data •
Statistical tests for association and
difference • Presenting the data – tables
and charts • Validating the data
⦿ Advantages of quantitative analysis •
Disadvantages of quantitative analysis •
Checklist for the use of basic statistics
 Sources of quantitative data
⦿ Quantitative data take the form of numbers.

⦿ They are associated primarily with (but not limited

to) strategies of research such as surveys and
experiments, and with research methods such as
questionnaires and observation
Types of quantitative data
⦿ Nominal data; come from counting things
and placing them into a category. E.g.,
male/female or White/South Asian/African
Caribbean. No underlying order to the
names.
⦿ Ordinal data are based on counts of things
assigned to specific categories and the
categories stand in some clear, ordered,
ranked relationship which allows
comparison. E.g., the Likert scale below:
⦿ Interval data; categories are ranked on a scale; the
‘distance’ between the categories is a known factor
and that it is proportionate. E.g., 1966, 1976, 1986

⦿ Ratio data; the categories exist on a scale which has

a ‘true zero’ or an absolute reference point;
incomes, distances and weights

⦿ Discrete data; whole units (1,2,3..). Numbers of

children per family. We might aggregate the numbers
to arrive at an average figure of 1.9 or whatever, but
we do not suppose that there exists anywhere 0.9 of
a child belonging to a family.

⦿ Continuous data; measured ‘to the nearest unit’

because they do not come in discrete chunks. E.g.,
height in mm, weight in grams, age in seconds.
 Preparing quantitative data for
analysis
Coding the data; entails the attribution of a
number to a piece of data, or group of data;
advisable to do it prior to collection of data.

⦿ For example, instead of asking 1000 people to

state what their job is, they can be asked to
identify which category of occupations best fits
their own.

⦿ The result of doing it this way is that, if there

are, say, ten categories of occupation used, the
researcher will obtain results in the form of
numbers in each of the ten categories, rather
than the list of 1,000 words.
 Grouping the data (1st stage)

⦿ Organize the raw data in a way that makes

them more easily understood. As Table 13.2
shows, it is difficult to make sense of raw
data when the no. of items goes above ten.
You should construct an array of the raw data,
i.e. to arrange the data in order.

Make a tally of the frequencies

⦿ If there are a large no. of frequencies, make
grouped frequency distributions
 Describing the mid-point or average
(a measure of central tendency)

1) The mean (the arithmetic average); It is based on

equal distribution of values. Add all the values and
divide by the no. of cases. E.g., if the total =116, and
the no. of cases is 8. 116/8= 14.5.
2)The median (the middle point); Values in the data are
placed in either ascending or descending rank order
and the point which lies in the middle of the range is
the median.
In the following set of data, the median is 11.5, half-way
between the two middle values.
3) The mode (the most common); value which
is most common. E.g., 17 in the set below.
 Describing the spread of data
(a measure of dispersion)

⦿ 1) The range; subtract the minimum value

from the maximum value. In the set below,
the range is 44 (i.e. 47 minus 3).

⦿ It may give a biased picture of the spread of

the values between the extremes. Here, the
range of 44 is determined by the outlier
value of 47.
2) Fractiles; (IQR) Fractiles are based on the idea of
dividing the range of the data into fractions with
each fraction containing the same number of data
cases.
When quartiles are used, the data are divided into
four sections. For deciles, you divide the range into
tenths. For percentiles,100 equal parts.
3) Standard deviation; To understand the dispersion
of the data, it would be useful to know how far, on
average, the values varied from the mean.
Looking for patterns and relationships
(looking for connections between variables)
⦿ Does one category of data match another, or vary
along with another? Is there a difference between
two sets of data where similarity was expected?

⦿ Find out: •whether the findings were a fluke; •how

strong the connection is between the two variables;
•whether one causes the other, or whether they are
mutually interdependent

⦿ Start with a ‘null hypothesis’; the presumption that

there is no real relationship between sets of data.
⦿ Only if the patterns or relationships are shown to be
statistically significant (p < 0.05) should you be
persuaded to reject the null hypothesis.
Statistical tests for association
and difference
How do I find out if two variables are
associated to a significant level?

⦿ Probably the most flexible and certainly the

most commonly used statistical test for this
is the chi-square test. It works with nominal,
ordinal, interval and ratio data
⦿ It works on the supposition that if there were
no relationship between the variables then
the units would be distributed in the cells of
the contingency table in equal proportion.

⦿ This means that if there were three times as

many women as men in a sample, we would
start from the assumption that women would
account for 75 per cent of the amounts or
frequencies
How do I see if two groups or categories are different
to a significant level?
⦿ A teacher might wish to compare the results obtained
from students in one class with those from another.

⦿ The best statistical test for this purpose is the t-test.

It uses the means of the two sets of data and their
standard deviations to arrive at a figure which tells
you the specific likelihood that any differences
between the two sets of data are due to chance.

⦿ When using the t-test, take the null hypothesis as the

starting point. Treat differences as ‘real’ when you
find that there is a probability of ‘p < 0.05’, that any
difference between the two sets of data were due to
chance.
Example: The t-test
⦿ We have two sets of marks obtained from two
classes who have both taken the same test at
about the same time. Is there a significant
difference between the scores obtained by
students in class A and the scores obtained by
those in class B?

Works well with small sample

sizes (less than 30) and the
groups do not have to
be exactly the same size.
How do I see if three or more groups or
categories are significantly related?

⦿ It suggests a fairly large data set with more

than two variables potentially connected and
worthy of statistical test.

⦿ It calls for a basic factor analysis, such as

one-way ANOVA (analysis of variance). This
test analyses the variation within and
between groups or categories of data using a
comparison of means.
How do I assess the strength of a relationship
between two variables?
⦿ Correlations between two variables can be
visualized using a scatter plot and they need
ratio, interval or ordinal data, and cannot be
used with nominal data.
⦿ They also require reasonably large data sets
⦿ Spearman’s rank correlation coefficient
(which works with ordinal data) and
Pearson’s product moment correlation
coefficient (which works with interval and
ratio data) are the statistical tests used to
arrive at a correlation coefficient based on
the data.

⦿ 0.3 is reasonably weak, 0.7 is reasonably

strong.
 Presenting the data – tables
and charts
Vital information
The table or chart must always have:
• a title;
• information about the units being represented
• the source of the data, if they were originally
produced elsewhere.

• Horizontal axis (X) - independent variable;

•the vertical axis (Y) - dependent variable.
 Tables

contingency tables present a comparison of sets

of nominal data.
 Bar charts
⦿ Used with both nominal and discrete data. The
bars should be of equal width, with the height of
the bars representing the frequency or the
amount for each separate category.
 Histograms
⦿ A histogram is used for continuous data,
whereas a bar chart is used for discrete
data or nominal data: •no gaps between
the bars; • the data ‘flow’ along the x
axis, rather than being separate items.
 Scatter plots
⦿ used to display the extent of a relationship
between two variables.
⦿ The closer the points come together, the closer
the relationship between the variables. Good at
showing patterns and deviations
 Line graphs
⦿ Used for depicting development or
progression in a sequence of data. Good for
showing trends in data.
 Pie charts
⦿ They convey the proportions of each category which
make up the total. Mostly, the segments are
presented in terms of percentages. No more than
seven segments and no segment which accounts for
less than 2 % of the total.
 Validating the data
⦿ The research instrument does not vary in the
results it produces on different occasions when it
is used.

⦿ Respondents provide the same kind of answers to

similar questions and generally answer in a
consistent fashion.

⦿ Different researchers would arrive at similar

conclusions when looking at the same data.

⦿ The findings will apply to other people and other

contexts.

⦿ The analysis is ‘correct’ - ‘works’’

 Advantages of quantitative
analysis
⦿ Scientific; statistical techniques based on the
principles of mathematics and probability.

⦿ Confidence; Statistical tests of significance give

researchers additional credibility

⦿ Measurement; Interpretations and findings are

based on measured quantities rather than
impressions

⦿ Analysis; Large volumes of quantitative data can

be analysed relatively quickly

⦿ Presentation; Tables and charts provide a

succinct and effective way of organizing
 Disadvantages of quantitative
analysis
⦿ Quality of data are only as good as the methods
used to collect them and the questions that are
asked.

⦿ Technicist. There is a danger of researchers

becoming obsessed with the techniques of
analysis

⦿ Data overload; Too many cases, variables,

factors to consider – can cause complexity.

⦿ Quantitative analysis is not as scientifically

objective as it might seem on the surface.