See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/353828687

Statistical tests and Measurements

Chapter · July 2017

CITATIONS READS

0 7,275

2 authors:

Huma Parveen Mubashir Gull

Aligarh Muslim University GITAM University
25 PUBLICATIONS 87 CITATIONS 27 PUBLICATIONS 48 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Huma Parveen on 11 August 2021.

The user has requested enhancement of the downloaded file.

 Quadrant-I (e-Text)

Module name/ title: Statistical Tests & Measurements

Paper: Communications Research

Component I: Personal Details

Role Name Affiliation

Principal Investigator Prof. Biswajit Das Centre for Culture,
Media & Governance,
JamiaMilliaIslamia,
Delhi
Co-Principal Dr. Durgesh Tripathi University School of
Investigator Mass Communication,
Guru Gobind Singh
Indraprastha University,
Delhi
Paper Coordinator (if Dr. Sunitha Chitrapu Social Communications
any) Media Department
(SCMSophia), Sophia-
Smt. Manorama Devi
Somani College,
Mumbai 400 026
Content Writer/ Dr. Huma Parveen Dept. of Mass-
Author(s) Communication, Aligarh
Muslim University
Mubashir Gull
Department of
Psychology, Aligarh
Muslim University,
Aligarh
Content Reviewer Prof. Biswajit Das Centre for Culture,
Media & Governance,
Jamia Millia Islamia,
Delhi
Language Editor Mr. P K Satapathy Department of English,
School of Open
Learning, University of
Delhi

Component II: Description of the Module

Items Description of Module

Subject Name Communication Studies
Paper Name Communications Research
Module Name/Title Statistical Tests & Measurements
Module ID P2M20

Media & M20 - Statistical tests and Measurements 1

Communication
Studies
P2 – Communications Research
 Quadrant-I (e-Text)

Pre-requisites
Objectives
Keywords t-test; z-test; Anova; chi square test

Module 20: Statistical Tests & Measurements

1. Introduction
Mathematical science dealing with the collection, analysis, interpretation, and
presentation of numerical data in order to draw appropriate conclusionsis called
statistics (Trajkovski 2016).It is an independent branch and its use is highly prevalent
in all the fields of knowledge. Methods such as parametric and non-parametric tests
are used in statistics. Statistics is used both in scientific and non-scientific way to
make appropriate decisions and conclusions based on the data.

Statistics which are based on the normal distribution of the data are called parametric
statistics. T, z, and f are examples of parametric statisticaltests. Those statistical
tests, which are not based on normal distribution of the data, are called non-
parametric statistics or distribution free tests.Chi-square test,Spearman’s rank
correlation coefficient, Mann-Whitney-U test, Kruskal-Wallis analysis of variance etc
are some of the examples of non parametric tests. Furthermore, parametric tests are
more powerful statistical tests than non-parametric tests. It is always recommended
to use parametric statistical tests than non-parametric test. However, non-parametric
statistical tests are only used when the assumptions of the parametric tests are not
met or fulfilled.

2. Assumptions of Parametric and Non Parametric tests

2.1. Assumptions of Parametric tests:

 The distribution of the sample should be normally distributed.
 The sample is in the form of interval or ratio scale
 The observation must be independent.
 The populations must have the same variance
 The samples must have equal or nearly equal variances and

2.2. Assumptions of Non Parametric tests:

When the assumptions and conditions of the parametric statistical procedures are
not met, in such a situation we are bound to use non-parametric statistics. Non
parametric statistics are used in such a situation:
 If the sample is not normally distributed
 If our sample data has been measured at nominal or ordinal scale
 The sample size is very small.

Trajkovski, (2016) in his research paper mentioned that wrong statistical tests are
used by the researchers. He reported that most of the researchers use parametric
test which does not fulfill the assumptions of parametric tests and vice versa. Now
days, the availability of the different types of statistical software’s like SPSS, AMOS,
MATLAB, R software, etc. makes performing of the statistical test easy but selection
of the appropriate statistical test is still a problem. A systematic step-by-step
approach is the best way to decide how to analyze the data. Hence it is
recommended to follow the below mentioned steps before analyzing the data.

Media & M20 - Statistical tests and Measurements 2

Communication
Studies
P2 – Communications Research
 Quadrant-I (e-Text)

1) Identify the research question.

2) Frame the question in the form of a statistical hypothesis such as the null
hypothesis (no relationship between the variables) and alternative hypothesis
(positive or negative relationship between the variables).
3) Determine whether the variable is dependent variable or independent
variable.
4) Use appropriate research design, which controls the confounding variables.
5) Use the appropriate statistical test.
6) Do a power analysis to determine a good sample size.
7) Collect the data.
8) Examine the data to see whether it meets the assumption of parametric or a
more appropriate substitute test.
9) After this, apply the statistical test you chose, and interpret the results.
10) Show your results effectively, usually with a table or a figure.

3. T-test

A statistical test that offers an opportunity to compare between two group means is
called t-test. For example, if we want to compare between male and femaleon health
issues or if we want to compare between rural and urban people on the same, t-test
is the appropriate statistical test to measure the difference between these
groups.Other statistical tests include an z-test, chi-square test and analysis of
variance. T-test is used when we have smaller group of data while as z-test is used
when we greater group of data (>30).It verifies, if the difference between two means
is larger than would be expected by chance. Common types of t-tests which are
frequently used are dependent sample t-test and independent sample t-test.

i. Dependent samples t-test

A statistical test which is used to compare between two sample means on a single
variable (dependent variable) is called dependent sample t-test. In this, the
participants are meaningfully related with each other as the participants are same in
both pre and post test. Unlike independent sample t-test, dependent sample t-test is
used to compare the mean of a single sample or paired samples.

For example, if a researcher is interested to compare the mental health of the

participants before and after the national disaster such as (earth quake, tsunami
etc.). He will first administer the mental health inventory on a group of participants
before the national disaster. The same group of participants then participates in the
study after the tsunami. Then the researcher can compare the two means. In the
dependent sample t-test, each score is matched and because of this matching, the
researcher can predict that the scores are interdependent. For such kind of study,
dependent sample t-test is the appropriate statistics.

ii. Independent samples t-test

Independent sample t-test is used when we have to make a comparison between two
sample meanswho’s means are not dependent on each other. Unlike dependent
sample t-test (where the participants are meaningfully related with each other), two
separate groups of participants participate in the study. One of the commonly used t-
test is the independent sample t-test, where each groups are completely independent

Media & M20 - Statistical tests and Measurements 3

Communication
Studies
P2 – Communications Research
 Quadrant-I (e-Text)

of each other. Comparision between male and female, Urban & rural areas is the
simple examples of independent sample t-test. If we want to make a comparision
between male and females on mental health, an independent sample t-test is the
appropriate statistics. In this test, the sample of men should not be related to the
smaple of females, and the should not be any kind of overlape between the two
groups, i.e the groups should be independent of each other.

3.1. Assumptions of t-test:

The common assumptions made when doing a t-test include:
 The scale of measurement should be continuous or ordinal in nature
 The data should be selected through probability sampling technique
 The data should be normally distributed and
 The sample should be adequate.

4. Z-test
Like t-test, z-test is a statistical test that offers an opportunity between the two groups
but unlike t-test, where the variance is unknown and sample size is small, in z-test,
there is known variance and sample sizelarger.For example, if we want to compare
between male and female on mental health, z-test is the appropriate statistical test to
measure the difference between thegroups. The condition is that, the data should be
greater than thirty and normally distributed, and the standard deviation should be
known. Besides that while conducting a z-test, the null and alternative hypotheses,
alpha and z-score should be stated.

4.1. Assumptions of z-test:

 The sample must be independent of one another
 The samples must be normally distributed
 The sample sizes must be large (≥30) and
 The standard deviation and population variance should be known.

5. Analysis of variance (ANOVA)

While as the t-test and z-tests are used when we have to compare the differences
between two groups. The problem arises when we have to compare more than two
groups. In such a situation when we to compare among three or more than three
groups, Analysis of variance (ANOVA) is an appropriate statistics. It is nothing, but
an extension of t-test and z-test. It is better to say the technique as analysis of
means rather than variance as inference about the means are made by analyzing the
variances. This test is used to test general rather than specific differences among
means. While performing analysis of variance, we get two variances, between group
variance and within group variance. Difference between the means is called between
group variance and difference within the means is called within group variance.

If the difference between the two group variances (i.e. between and within groups) is
significant i.e., between group have large variance as compared to within group, we
reject the null hypothesis and conclude that our experimental manipulation had a real
effect. If the difference between the variance is not significant, we accept the null
hypothesis and conclude that experimental manipulation didn’t have real effect.
Unlike t-test, where we get the t value, In analysis of variance, we get the Fratio. The
F ratio is computed by dividing the between groups variance estimate by the within
group variance.

Media & M20 - Statistical tests and Measurements 4

Communication
Studies
P2 – Communications Research
 Quadrant-I (e-Text)

There are different types of ANOVAs such as (one way ANOVA randomized, one
way ANOVA repeated, two way ANOVA randomized, two way ANOVA repeated and
factorial ANOVA)

A one way ANOVA is used whenever we have one independent variable with three
or more than three levels. If separate groups of subjects participate in each
condition/level, in a between subject design, a randomizedANOVA is the appropriate
test. However if the same subjects have participated in each condition, in a within
subject design, a repeated measure ANOVA is the appropriate test. While as a two
way ANOVA is used whenever two independent variables are manipulated and all
the combinations of levels of each of the two variables are used.

When we have one continuous dependent and two or more categorical independent
variables, the factorial anova is the appropriate statistics. Besides mean difference,
factorial anova provides the main as well as the interaction effect.

For example, I want to know the difference between the boys and girls on academic
performance in school. Further, I also want to know the difference between the
children’s coming from rural and urban areas on academic performance. In this
example, the academic performance is the dependent variable, and gender of the
children and area of location (rural/urban) are two independent variables. This is
known as 2 (gender)× 2 (location) factorial analyses.

5.1. Assumptions of Analysis of Variance (ANOVA)

The common assumptions made when doing a ANOVA include:
 The sample should be drawn from a normally distributed population
 All populations should have a common variance.
 All samples should be drawn independently of each other.
 Within each sample, the observations are sampled randomly and
independently of each other

6. Difference between t, z and f test

Z-test is used to compare the mean of two groups, with large sample size whether
population standard deviation is known or not. While as t-test is used to compare
between the two groups when the population standard deviation is not known. An F-
test is an extension of t-test and z-test and is used to compare mean score of more
than two groups and population variance of any sample size.

7. Chi-Square Test

Test which is used to determine the difference between expected frequencies and
observed frequencies in one or more than one categories is called chi-square test. It
is one of the most commonly used non-parametric tests used by social science
researchers.The Chi square test is a statistical testwhich measures the association
between twocategorical variables (Ugona & Walker 1995).
It is a computationally simple statistical test which is used to examine independence
across two categorical variables.

Media & M20 - Statistical tests and Measurements 5

Communication
Studies
P2 – Communications Research
 Quadrant-I (e-Text)

Chi Square performs two important functions

a. Goodness of fit
A statistical model which is used to describes how well it fits the observations. It is
used to compare the observed values with the expected value. In this test, the data is
first divided into intervals and then points that fall in the intervals are compared with
the expected points in each interval.

b. Measure of Independence

One of the most useful statistics for testing the hypothesis, when the data are
nominal is the chi-square test of independence. It is used to determine the
significance of association between the two variables (categorical variables) when
the sample size is large. This test is used when we have to compare two nominal or
categorical variables and we want to know the difference in terms of proportion
between the variables. For example, chi-square test of independence is used to
determine whether gender (Male, Female) is related to voting preference (Congress,
BJP, Independent etc.).

According to Kothari (2007), chi-square test is simply a technique of association

between two attributes. It evaluates the probability that the observed relationship
results from chance (Koul, 1984). If the calculated value of chi-square test is less
than the tabulated value, we could conclude that the two variables are not related
with each other. If however the calculated value is greater than the tabulated value,
then we could conclude that there is a significant association between the variables
(Onchiri, 2013).

7.1. Assumptions of Chi-Square Test

 The data should be Quantitative in nature.
 It should have one or more categories.
 Each observation should be Independent.
 The sample should be adequate in size and no cell should have an expected
frequency of less than 5.
 Data should be in frequency form and
 All the observations must be used.

8. Summary
In this paper, we have described the basics of a parametric tests (t-test, z-test and
ANOVA) and non parametric tests (chi square test). We hope that information
provided has clarified the difference between parametric and non parametric
statistics. In addition, we hope that information provided has clarified the difference
between independent sample t-test and dependent sample t-test; one way ANOVA,
two ways ANOVA, factorial ANOVA; and chi-square test. The paper finally describes
the assumptions of all the above mentioned statistical tests.

References
1. Hulsizer, M.R., & Woolf, M. L. (2009). A Guide to Teaching Statistics:
Innovations and Best Practices, United Kindom:A John Wiley & Sons, Ltd.
2. Kothari, C. R. (2007). Quantitative techniques. New Delhi, UBS Publishers

Media & M20 - Statistical tests and Measurements 6

Communication
Studies
P2 – Communications Research
 Quadrant-I (e-Text)

3. Koul, L. (1984). Methodologies of educational research. New Delhi, Vikas.

4. Mangal, S. K. (2006). Statistics in Psychology and Education. New Delhi:
Prentice Hall of India.
5. Onchiri, S. (2013). Conceptual model on application of chi-square test in
education and social sciences. Education research and reviews, 8(15), 1231-
1241.
6. Singh, Y. K. (2006). Fundamental of Research Methodology and Statistics.
New Delhi: New age international publisher.
7. Trajkovski, V. (2016). How to Select Appropriate Statistical Test in Scientific
Articles. Journal of Special Education and Rehabilitation, 17(3–4)5–28.

Media & M20 - Statistical tests and Measurements 7

Communication
Studies
P2 – Communications Research
View publication stats