Week 11 Tutorial

In this session you will learn how to conduct a parametric inferential statistic that
allows you to test for differences in scores across either two levels (paired
sample t-test) and its non-parametric equivalent and three or more levels of your
independent variable simultaneously i.e. three or more conditions in your
experiment (one-way repeated measures ANOVA) and its non-parametric
equivalent. This week, we focus on repeated-measures designs i.e. one group of
participants tested under multiple conditions on the same task.

Task: Paired t-tests

A clinical psychologist is attempting to assess the effectiveness of a new informal

intervention to alleviate symptoms of depression amongst his colleagues. He measures
theirsymptoms via the BDI (Beck Depression Inventory) at two time points: before and after
the intervention – the higher the score the higher the depression. The scores are below:

Beck Depression
Inventory scores
Pre- Post-
intervention intervention
48 40

46 36

42 37

40 31

38 39

56 39

52 45

52 46

52 42

44 42

48 42

51 43

52 43

50 45

60 46

58 45
 Q. Sketch out (i.e., copy and paste from JASP) a histogram showing the
difference between the pairs ofscores. Is this normally distributed?

Q. What are the appropriate descriptive statistics to report? Compute

these in JASP and report them in a table below.

Q. Conduct the appropriate inferential statistic and copy into the

table below.Make sure you highlight the relevant sections.

Paired Samples Test

Paired Differences
95% Confidence Interval of
Std. Std. theDifference Sig.
Mean t df (2-tailed)
Dev Error Lower Upper
Mean
Pair
1

Q. How are the degrees of freedom derived?

Q. What do the 95% CI mean?

Q. Calculate the effect size (Cohen's d). Interpret this:

(mean 1 – mean 2)
SD of the difference
 Q. Write up the findings as you would find in a published journal
article. Make sure you include these details: recap of experiment,
the design, descriptives, conventional statistical evidence, effect size
and confidence intervals.

Task. Wilcoxon signed rank test.

Practice effects can be a confounding variable when measuring cognitive performance. To rule out that such
effects may exist, an experimenter was carried out on the number of errors made in the Wisconsin Card
Sorting Task. 32 participants were measured at baseline (time 1) and one day after (time 2). Fewer errors
indicate better performance. Because the data did not meet various assumptions of parametric testing for a
paired sample t-test, the researcher conducted a Wilcoxon signed- rank test instead. Conduct the test and
report/interpret the results. When reporting the results, include the median (run descriptive stats for this) and
effect size.

Q. Some values are missing from this table. See if you can if you can
calculate what these values are:

Within Subjects Effects

Sum of Squares df Mean Square F p partial η²

Time Bin 2 13.03 .014 .30

Residual 60.62 24

Note. Type 3 Sums of Squares

Q. How are the two df calculated?

Q. What does the effect size (ηP2) SUGGEST for this study?
1
Again you can also do it with just two conditions instead of a paired t-test, and in that case the resultant statistic is
directly comparable (i.e. F = t 2)

Task. One-way repeated measures ANOVA

Researchers wished to research the effect of MDMA (ecstasy) intake upon physical
endurance training on new teenage recruits to the cadet forces. Participants were tested
(randomly) in four different conditions with varying doses; at 0mg/kg (baseline), 20mg/kg,
50mg/kg and 100mg/kg. Recruits were required to run in full battle gear until they themselves
asked to stop. The distance travelled (km) was recorded. The data obtained is shown below:

0g/k 20mg/k 50mg/k 100mg/k

g g g g
8.2 8.6 9.7 10.2

9.3 10.3 12.9 11.8

6.5 12.5 11.6 11.9

7.2 15.3 11.2 12.0

8.3 7.8 10.8 10.5

10.2 9.6 9.8 12.6

7.6 10.2 8.9 9.6

7.4 8.5 10.7 10.9

6.9 8.9 13.2 12.5

8.3 9.6 12.3 11.0

9.7 9.9 12.5 14.2

5.9 11.0 14.0 13.0

7.8 8.3 9.3 13.1

6.2 7.9 8.4 9.9
Q. Conduct the statistical analysis and complete the table below:

Within Subjects Effects

Sum of Squares df Mean Square F p partial η²

Dosage
Residual

Note. Type 3 Sums of Squares

Q. When might you report the corrected values (i.e. Greenhouse-Geisser

or Huynh-Feldt) instead?

Q. What may you conclude from the analysis so far?

Q. Write up this entire study:

Task. Friedman’s ANOVA.

There is anecdotal evidence that using Facebook and Instagram for extended periods of time can make you
dumber. To test this hypothesis, a group of researchers recruited 54 participants, and in a counterbalanced
fashion, got them to spend 2 hours using Facebook, 2 hours using Instagram and 2 hours playing Chess over
three different days. At the end of each session, participants completed a short IQ test, and performance was
compared across the three conditions. Because the data did not meet various assumptions of parametric testing
for a one-way repeated measures ANOVA, the researcher conducted a Friedman’s ANOVA on the data.
Conduct the test and report/interpret the results. When reporting the results, include the effect sizes (i.e. for
the three comparisons).