STATISTICS FOR COMMUNICATION STUDY

CAUSAL-COMPARATIVE
RESEARCH

Haris Herdiansyah, S.Psi, M.Si

Communication Study Program

President University
2024
Survey/Descriptive
Research
 Causal-comparative research
• This type of research is to determine the
cause for, or consequences of, existing
differences in groups of individuals
• Also known as “ex post facto” research
• Prospective causal-comparative research
occurs when a researcher initiate a study
beginning with the causes and it’s
determined to investigate the effects of
a condition
 Example of causal-comparative research

• One researcher compared the levels of self-confidence between

students who came from vocational high schools and high schools
when they started university

Comparison the level of self-confidence between vocational high

school students and senior high school students when starting
university
 Another Example

• A researcher formed 3 groups of preschoolers; 1) those who never

watched educational YouTube channel, 2) those who watched it
sometimes, 3) those who watched it frequently.
• The researcher compared the 3 groups on a reading readiness test

The influence of educational YouTube channel toward

reading readiness test on preschoolers
Statistics tests for Causal-Comparative
Research
• Statistics are extensively used in causal comparative research such as:
Pearson Product-Moment Coefficient
Spearman Rank Order Coefficient
Phi Correlation Coefficient
Regression
t-tests
Chi-Square
Analysis of Variance
 What is a t-test ?

• A t-test is a statistic that checks if two means

(averages) are reliably different from each
other
• t-test is part of inferential statistics that allow
us to make inferences about the population
beyond our data
• Minimum requirement scale of data for t-test
is Interval data
 Types of t-test
• One sample t-test  test the mean of one group against a set mean
example : testing the students IQ comparing with the norm or the
average of 100
• Two sample t-test
 Independent sample t-test/between-sample t-test/unpaired sample t-
test  test the means of two different groups
example : testing the quality of two different cosmetic brand
 Paired sample t-test/within-subjects t-test/dependent-sample t-test
 tests the mean of one group for twice
Example : testing the power of remembering something before and
after drinking the brain supplement
Comparison Types of t-test
 Parametric VS non-Parametric in t-test

Parametric Requirements Purpose Non-parametric

t-test t-test
One sample t-test Normality test To identify the difference of Proportion test
sample mean with the
population mean/norm
Independent sample t- Normality test to find out whether there is Wilcoxon test
test a difference in the mean of
the two sample groups that
are separate or unpaired
from each other.
Paired sample t-test Normality test to find out whether there is Mann Whitney test
a difference in the mean
between two sample data
that are related to each
other
One Sample t-test
 One sample t-test

• The purpose of the One sample t-test is to compare the mean of the
sample under study with the mean of the existing population.
• One sample t-test is part of the parametric statistical test with the
basic assumptions that must be met, namely the research data must
be normally distributed, the research data is at least interval data.
Case 1
• According to a marketing manager's report at a
cellphone sales center, the average daily sales of
"oddo" brand mobile phones in all stores is 85 units.
• You, as a researcher, are interested in testing the
truth of the report, whether it is true that the daily
sales of Odo brand mobile phones are 85 units. For
that, you need to test it using a sample of 30 stores
at random (simple random sampling).
• The average daily sales data from 30 sample stores is
in the Microsoft excel file
 Research Question & Hypothesis
• RQ
Was the average daily sales of “oddo” brand mobile phones is 85 units
per day?
• Hypothesis
H0 = there is no difference in the average daily sales of "oddo" brand
mobile phones between the average in the population and the
average in the sample (= 85)
Ha = there is a difference in the average daily sales of "oddo" brand
mobile phones between the average in the population and the
average in the sample (≠ 85)
 Steps of Analysis

1. Conduct a normality test to find out whether or not the average

daily sales of 30 sample stores are normally distributed.
2. Conduct a one sample t-test using SPSS software to see if it’s true
that the daily sales of oddo brand mobile phones are 85 units per
day
• Open the SPSS Software, then make some
adjustment
Name  oddo
Label  average sales per day
• Move the raw data from excel to data view
• the following display will appear
Normality Test
• Click “Analize”  “Descriptive Statistics”
 “Explore”

• Move “average sales per day” to

Dependent list box
• Click “OK”
Output of Normality test

• There are two distinct columns; Kolmogorov-Smirnov and Shapiro Wilk.

• Using Shapiro-Wilk if the respondents involved or the amount of data analyzed ≤
50
• Using Kolmogorov-Smirnov if the respondent or the amount of data analyzed is >
50
• For this example, because the number of respondents and the data are 30, the
output of Shapiro-Wilk is used
• To find out whether or not the selected sample stores is normally
distributed, we need to look at the sig score.
• If the sig value > 0.05 then the selected sample is normally distributed
• If the value of sig < 0.05 then the selected sample is not normally
distributed
• From the SPSS output above, the value of sig. is 0.138 (> 0.05) then the
selected sample is normally distributed and the prerequisites for
conducting the t-test have been met
 t-test

• Click “Analize”  “Compare Means” 

“One-Sample T-Test”
• Move “Average sales per day” to Test
Variable (s) box
• Write 85 in the test values as the
population mean of this case example
• Then click “OK”
 Result

• To find out the result, whether there is a difference in the average daily sales
between the sample and the population, it can be seen from the sig score. (2-tailed)
compared to 0.05
• If the score sig. (2-tailed) > 0.05 there is no difference (H0 accepted, Ha rejected)
• If the score sig. (2-tailed) < 0.05 there is a difference (Ha accepted, H0 rejected)
• The results show a score of sig. (2-tailed) = 0.004 (< 0.05)
• It means; there is a difference in the average daily sales between the sample and the
population
• H0 is rejected, Ha is accepted
Independent Sample t-test
Independent sample t-test
• Independent sample t-test is used to determine whether there is a
difference in the mean of two sample groups that are separate or
unpaired from each other.
• The requirements for independent sample t-test are: test for
normality
Case 2
• During the pandemic, the university implemented two learning
models; online and offline. A statistics lecturer wants to know the
difference in statistical learning outcomes between online and offline
learning models. Class A represents the online learning model, and
class B represents the offline learning model.
• Class A and B learning outcomes data are in the Microsoft excel file
Research question & Hypothesis
• RQ
Is there a difference in statistical learning outcomes between class A
using the online learning model and class B using the offline learning
model?
• Hypothesis
H0 = There is no difference in statistical learning outcomes between the
online learning model and the offline learning model
Ha = There is a difference in statistical learning outcomes between the
online learning model and the offline learning model
Steps of analysis
• Click “Variable view” and make
adjustment
• For Values column  type “1” for
“Online model” and”2” for “Offline
model
• Click “OK”
• Enter all data in excel to data view.
• In the outcome column, enter the
statistical value data for all
students, and in the model column
enter the code "1" for online
classes and "2" for offline classes.
 Normality test
• Click “analize”  “descriptive statistics”
 “explore”
• Move “statistics learning outcome” to
Dependent list box and “model of
learning” to Factor list
• Click “plots”  click “Normality test
with plots”  “continue”
• Click “OK”
Output of normality test

• We use the Shapiro-wilk because the data size is less than 50 data
• If the sig value > 0.05 then the selected sample is normally distributed
• If the value of sig < 0.05 then the selected sample is not normally
distributed
• Both online and offline model are normally distributed because all the
score are > 0.05
Independent sample t-test
• Click “Analize”  “Compare mean” 
“Independent sample t-test”
• Move “statistics Learning Outcome” to Test
variable(s) box and “Model” to Grouping
variable box
• Click “define group”
• Put “1” for Group 1 and “2” for Group 2 
“continue”
• Click “OK”
 Output of Independent sample t-test

• First, look at sig. of Levene’s Test Equality of Variances

• If the score of sig < 0.05 use Equal variances not assumed, otherwise if sig>
0.05 use Equal variances assumed
• In this case the sig. score is 0.786 > 0.05, then use the Equal variances not
assumed to find out the result of Independent sample t-test
• Second, look at Sig. (2-tailed) as a basis for determining whether or not
there is a difference between the online and offline models (whether H0 is
accepted or Ha is accepted)
 Output of Independent sample t-test

• If sig. (2-tailed) > 0.05  there is no difference …. (H0 accepted, Ha rejected)

• If sig. (2-tailed) < 0.05  there is a difference …. (Ha accepted, H0 rejected)

• Based on the output above, the score of sig.(2-tailed) is 0.051 (> 0.05) it means that
there is no difference in Statistics learning outcome between online learning model
and offline learning model  H0 accepted, Ha rejected
Paired Sample t-test
Paired sample t-test
• Is a comparison test that aims to find out whether there is a
difference in the mean between two sample data that are related to
each other
• The paired sample t-test has a precondition that the distribution of all
data must be normal so it is necessary to test for normality.
Case 3
• Because the outcome of statistics courses is always low, the lecturer think
a special training is needed to improve the understanding and outcome of
statistics for communication students. For this reason, special training in
statistics courses was given .
• To find out whether the special training has an impact to students, it is
necessary to measure the students' outcome before and after the
training.
• lecturer want to know whether there are differences in the statistical
outcome of communication students before and after being given special
training
• Pretest and posttest data on learning outcomes can be seen in the excel
file
Research Question & Hypothesis
• RQ
Is there any differences in the statistical outcome of communication
students before and after given special training?
• Hypothesis
H0 = there is no difference statistical outcome of communication students
before and after given special training (there is no influence of special
training toward statistical outcome of communication students)
Ha = there is a difference statistical outcome of communication students
before and after given special training (there is an influence of special
training toward statistical outcome of communication students)
 Steps of analysis

• Open the SPSS software  click

“Variable view” and make adjustment.
• Name  Pretest and Posttest
• Move the raw data from excel to data
view
Normality test
• Click “Analize”  “Descriptive statistic” 
“explore”
• Move “pretest” and “posttest” to Dependent
List box
• Click “Plots” then make adjustment
• click “Normality plots with tests” 
“continue”  “OK”
Output of Normality test

• We use the Shapiro-wilk because the data size is less than 50 data
• If the sig value > 0.05  the selected sample is normally distributed
• If the value of sig < 0.05  the selected sample is not normally distributed
• From the output above we can read that those data are normally
distributed because all the score are 0.334 and 0.245 (> 0.05)
Paired sample t-test
• Click “Analize”  “compare mean” 
“Paired sample t-test”
• Move “PreTest” and “PostTest” to
Paired Variables box
• Click “OK”
 Output of Paired sample t-test

• If sig. (2-tailed) > 0.05  there is no difference …. (H0 accepted, Ha rejected)

• If sig. (2-tailed) < 0.05  there is a difference …. (Ha accepted, H0 rejected)

• Based on the output above, the score of sig.(2-tailed) is 0.000 (< 0.05) it means that
there is a difference in Statistics learning outcome of communication students before
and after given special training (there is an influence of special training toward
statistical outcome of communication students)
• Ha accepted, H0 rejected
Lampiran
Thank You