Scribd
Upload
50 views

Independent Samples T-Test

Independent sample t-test

Uploaded by

Samiha Mjahed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
50 views

Independent Samples T-Test

Independent sample t-test

Uploaded by

Samiha Mjahed
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 30

Dr Daniel Boduszek

 Introduction to t-tests
 Types
 Assumptions
 The independent samples t-test
 SPSS procedure
 Interpretation of SPSS output
 Presenting results
 Paired samples t-test
 SPSS procedure
 Interpretation of SPSS output
 Presenting results
 If you open a book chapter on t-tests you will find
good information but you’ll encounter two things:

 The life story of the man who invented the t-test


& how he worked in brewery
 Complex mathematical formulae for how to
calculate t-tests by hand

• We will stick to what you actually need to know!


 Two main types:

 Independent samples t-test – compares


the mean scores of two different groups of
people or conditions

 Paired samples t-test – compares the mean


scores for the same group of people on two
different occasions

 Warning! Many different names for each type


• Compares mean scores collected from two
different groups of people.
• If we compared mean psychopathy scores
obtained from a group of prisoners to the mean
scores of lecturers we would have a between-
groups design.
• Comparing levels of criminal thinking between
violent & non-violent criminals is another
example of a between-groups design.
• Can you think of any other examples?
• Compares the mean scores & determines whether
they are significantly different.
• Tells you what the chance is that those two mean
scores came from the same underlying population.
• Do violent & non-violent criminals represent two
distinct populations when it comes to their styles of
thinking? Or are they both drawn from the same
population?
• A statistically significant
independent samples t-test (P < .05)
result tells us that it is extremely
unlikely that the two compared
scores are drawn from the same
population.

• We can, therefore, conclude that


these scores represent distinct
populations.
 The data are measured on a continuous (interval
scale)
 Independence of observations – observations must not be
influenced by any other observation (e.g. behaviour of each member of the group
influences all other group members)

 Normal distribution
 Random Sample (difficult in real-life research)
 Homogeneity of Variance – variability of scores for each of the
groups is similar.
 Levene’s test for equality of variances.
 You want non-significant result (Sig. greater than .05)
 Is there a significant difference in
the mean criminal behaviour scores
for violent and non-violent
offenders?
 From the menu click on
Analyze

 then select Compare


means

 then Independent
Samples T test
Move continuous DV
(recidivism) into the
Test variable box

& categorical IV
(type of criminal) into
Grouping variable box
 Click on Define groups
& type in the numbers
used in data set to code
each group

 Group 1 = 1
 Group 2 = 2

 Click on Continue

 & OK
 Checking the information about groups:

 Means
 Standard Deviations
 Number of participants in each group

Group Statistics

Type of Criminals N Mean Std. Deviation Std. Error Mean

Level of Recidivism 1.00 NonV 45 2.7556 1.73409 .25850

2.00 Violant 44 4.0000 3.32013 .50053


 Checking assumptions
 Levene’s test for equality of variance
 If Sig. value for Levene’s test > .05 – use the 1st line in the
table (Equal variance assumed)
 If Sig. value for Levene’s test < or = .05 – use the 2nd line in
the table (Equal variance not assumed)
Independent Samples Test

Levene's Test for Equality of


Variances t-test for Equality of Means

95% Confidence Interval of the

Mean Std. Error Difference

F Sig. t df Sig. (2-tailed) Difference Difference Lower Upper


Level of Recidivism Equal variances assumed 5.335 .023 -2.223 87 .029 -1.24444 .55969 -2.35690 -.13199

Equal variances not -2.209 64.513 .031 -1.24444 .56334 -2.36967 -.11921
assumed
 Differences between groups
 Check column Sig. (2-tailed)
 If Sig. value > .05 – no significant difference between
groups
 If Sig. value < or = .05 – significant difference between
groups
Independent Samples Test

Levene's Test for Equality of


Variances t-test for Equality of Means

95% Confidence Interval of the

Mean Std. Error Difference

F Sig. t df Sig. (2-tailed) Difference Difference Lower Upper


Level of Recidivism Equal variances assumed 5.335 .023 -2.223 87 .029 -1.24444 .55969 -2.35690 -.13199

Equal variances not -2.209 64.513 .031 -1.24444 .56334 -2.36967 -.11921
assumed
 The formula is:
 t2
 Eta squared = -------------------------
 t2 + (N1 + N2 -2)

 -2.212
 Eta squared = ---------------------------- = .05
 -2.212 + (45 + 44 -2)

 According to Cohen (1988)


 .01 = small effect
 .06 = medium effect
 .14 = large effect
 An independent samples t-test was conducted to
compare the recidivism scores of violent & non-
violent offenders. There was a significant difference
between the two groups, t(87) = -2.21, p < .05, with
violent offenders (M = 4.00, SD = 3.32) scoring
higher than non violent offenders (M = 2.76, SD =
3.32). The magnitude of the differences in the
means (mean difference = -1.24, 95% CI: -2.37 to -
.12) was small (eta squared = .05)
• Higher t scores equate directly to significant differences!
• The higher the t, the lower the P.
• A high t value tells you that the difference between the two
means is large, & so large that it is improbable (but not
impossible!) to have occurred due to chance.
• That is, sampling error – drawing your two mean scores from
separate ends of the one population distribution.
• If you understand this point, you understand just about
everything inferential statistics are about!
 A Paired samples t-test - works on very similar principles to the
independent t-test. However, it is used when you want to compare
two scores obtained from the same group of participants.
 Research question – Is there a significant change in prisoners’
criminal social identity scores after 2 year sentence in high security
prison? Does the process of prisonization have an impact on
prisoners’ criminal identity test scores? (example of a within-
subjects design).

 You need:
 1 categorical IV (Time 1, Time 2)
 1 continuous DV (criminal social identity test scores)
 From the menu
click on Analyze

 then select
Compare Means

 then Paired
Samples T test
 Click on the 2 variables that you are interested in
comparing for each subject (criminal identity,
criminal identity 2) & move them into Paired
Variables box

 Click OK
 Descriptive Statistics
Paired Samples Statistics

Mean N Std. Deviation Std. Error Mean

Pair 1 Criminal Identity 18.7303 89 8.93762 .94739

Criminal Identity2 26.3146 89 9.84031 1.04307

 Correlations
Paired Samples Correlations

N Correlation Sig.

Pair 1 Criminal Identity & Criminal 89 .941 .000


Identity2
 Differences between Time 1 & Time 2
 Check column Sig. (2-tailed)
 If Sig. value > .05 – no significant difference
 If Sig. value < or = .05 – significant difference

Paired Samples Test

Paired Differences
95% Confidence Interval of the
Difference

Mean Std. Deviation Std. Error Mean Lower Upper t df Sig. (2-tailed)

Pair 1 Criminal Identity - Criminal -7.58427 3.35006 .35511 -8.28997 -6.87857 -21.358 88 .000
Identity2
 The formula is:
 t2
 Eta squared = ----------------
 t2 + (N - 1)

 -21.362
 Eta squared = ------------------------ = .84
 -21.362 + (88 - 1)

 According to Cohen (1988)


 .01 = small effect
 .06 = medium effect
 .14 = large effect
 A paired samples t-test was conducted to evaluate
the impact of the prisonization process on prisoners’
scores on the criminal social identity. There was a
significant increase in criminal social identity scores
from Time 1 (M = 18.73, SD = 8.94) to Time 2 (M =
26.31, SD = 9.84), t(88) = -21.36, p < .001 (two-
tailed). The mean increase in criminal social identity
scores was -7.58 with a 95% confidence interval
ranging from -8.29 to -6.88. The eta squared
statistic (.84) indicated a large effect size.
• t-tests are used to compare two mean scores.
• When you are comparing mean scores obtained
from two different groups of people an independent
samples t-test should be used.
• When you are comparing mean scores obtained
from one group of people at two different time
points or under two different conditions a repeated
measures samples t-test should be used.
• A statistically significant independent t-test result
suggests that your two samples represent distinct
populations.
• A statistically significant paired t-test result
suggests that the change in mean scores is beyond
what should be expected by chance.
• The independent t-test makes a specific assumption
of homogeneity of variances. Therefore, the
Levene’s test results should always be inspected.
• What t-tests are used for
• The different types of t-tests available
• Independent samples t-test
• Paired samples t-test
• Research designs associated with each type of t-test
• Between-groups design
• Within-groups design
• The Levene’s tests
• The assumption of homogeneity of variance
Thank you for your
time!

You might also like