[L6] Analyzing Data from Independent Groups: Continuous and Ordinal Measures Flashcards
- Most common experimental design in psychological
research.
Independent Group Designs
Independent Group Designs- Used for comparing two (or more) groups which are
independent of one another – meaning there are ___
__
_ in each group.
different people
Independent Group Designs - Can be used for _____
true and quasi-experimental designs.
____ people are not assigned to
conditions, as they already belong to different groups.
Quasi-experimental designs –
Independent groups designs– __ different kinds of
__ can be used
three; dependent variable
Advantages of Independent Group Designs
Avoids the problems inherent in repeated measures
designs; practice effects, sensitization or carry-over
effects.
Disadvantage Independent Group Designs
more people, difficulty of matching
controlled group with experimental group
Statistical Tests for Independent Groups
Data
- The Independent Groups t-test
- The Mann-Whitney Test
- The Chi-Square Test
Use of test depends on data being tested.
Statistical Tests for Independent Groups
Data
The Mann-Whitney Test (also called the
Wilcoxon-
Mann Whitney Test)
also known as the between subjects t-test or the twosamples
t-test.
The Independent Groups t-test
Two (or Three) Assumptions about the Data
- Continuous/interval scale
- Data within each group is normally distributed. We
need to make sure that the data are approximately
normally distributed within each group. - The SD of the two groups are equal
there are actually two kinds of t-test, one makes
the assumption of ___, and one
does not.
homogeneity of variance
Common Mistake:
Normal Distribution and the
Independent Sample t-test
- To look at the overall distribution of data to determine
whether the data are appropriate for an independent
samples t-test. - It should be the ___ of the two groups.
distribution
Dispelling Myths
.
- Sample size must be above some value, such as 6, for
the t-test to be valid. - Sample sizes must be balanced, or similar
distribution of two (or more)
groups can be shown for the same amount of space.
Box and whisker plot –
We use the ____– approximately equal
sample sizes.
Pooled Variance t-test
Use ____ instead of histograms.
box and whisker plot
Check first the _____
(distribution of each group).
normal distribution assumption
The general form of the t-tests follow a
____
similar pattern.
Difference – in IG t-test, instead of being interested in
mean scores, we are now interested in the ____. So we call this ___
difference
between two means; d
The Equal Standard Deviations Assumption
* Also known as _____
Homogeneity of Variance.
Two versions of the t-test: the ____
which makes this assumption, and the _____ which does not make this assumption.
Pooled Variance t-test, Unpooled
Variance t-test,
If you have (approximately) equal sample sizes in your
two groups, use the ____
pooled variance t-test.
If you don not have (approximately) equal sample sizes
in the two groups, use the ____
unpooled variance t-test.
___ used to decide if variances (or SDs) are
the same. (most common)
Levene’s test
If Levene’s test give gives a ____
result, this means variances are different from one
another and the unpooled variance test should be used.
(vice versa)
statistically significant
Problem however is that a ____does not
mean that the variances are the same.
non-significant result
And the problems with the tests, such as the Levene test,
is that they are dependent on the _____
sample size.
When the sample size is small, it is _____ in the variances. (vice versa)
not very good at detecting differences
It only matters however that the variances are the same
when the sample size is ____.
small
So, when Levene’s test is good at telling us when the
variances are different precisely ____
we do not really care.
And when it is _____ when we do care.
not very good is precisely
Homogeneity of variance does not really matter when
the sample sizes are about ____
equal
So, if we have _____sample
sizes, we can ignore the assumption of Homogeneity of
Variance and use the Pooled Variances t-test.
equal (or approximately equal)
When the sample sizes are ___, Homogeneity of
Variance matters a lot more.
unequal
- A modification of the t-test, which does not make the
assumption of equal variances.
Unpooled Variance t-test
Sometimes known as ____which was developed
by ____
Welch’s Test ; Welch (1938)
Reasons for having Unequal Sample Sizes
- Comparison of two naturally occurring groups, and
they are different sizes. - It may be difficult or expensive for one of the
interventions. - There may be an ethical or recruitment issue.
We often describe the difference between two samples
by stating the
___
difference between the two means.
However, often the scale that is used is not one that
actually
____ (unless we are very familiar
with the scale)
makes any sense
How do we tell then if the difference is good or bad?
Use a measure of Effect Size,
in the case of the
independent groups t-test, an appropriate measure of
effect size is called __
_
Cohen’s d.
—= measure of how far apart the means of the
two samples are, in SD units.
Cohen’s d
It does not matter what the range of possible scores is –
we are interpreting it in terms of _
_.
SD
Cohen’s d is often interpreted according to the following
Rules:
- Large effect size: d = 0.8
- Medium effect size: d = 0.5
- Small effect size: d = 0.3
It can be as low as ____ and, unlike
a correlation (which can not be higher than 1), it has no
upper limit, although values above 1 are rare.
0 (It can not be negative)
*
* Used when independent samples t-test can not be used.
* Non-parametric test
The Mann-Whitney U Test
The Mann-Whitney U Test
Compares two unrelated groups.
The t-test compares the ___ of two groups and tells us
whether the
___ is statistically
significant.
means; difference in the means
The Mann-Whitney test does not compare means so it
might be tempting to say that it compares _
__.
Unfortunately it does not (necessarily)
medians
Effect Size for the Mann-Whitney test
* Use the value called ___, which is the Greek Letter ___.
ϴ; theta