Between Group Designs Flashcards
What do you need to have for between group designs?
2 separate groups!
In a between groups design, each participant only takes part in….
1 level of your IV
What is the difference between repeated measures and between group designs?
Between groups design- participants only take part in 1 level of IV
Repeated measures design- participants take part in all levels of IV
How should participants be assigned to a variable?
Randomly!
What is the purpose of random assignment to the groups?
Aim to evenly distribute random error.
Why is random assignment so important?
We don’t want systematic differences between our groups.
The maths assumes random error is spread equally across the 2 groups.
If you have parametric data and 2 separate groups- what statistical test are you going to use?
Independent t-test
The independent t test uses the t distribution to calculate the _____ of an observed event occurring.
Probability.
What’s important to consider about the scores of your groups?
To what extent they vary from each other.
If you have a low variability in group scores- what does this mean?
People have similar scores, aren’t varying as much, small SD.
Tiny bit of overlap between groups.
If you have medium variability in group scores- what does this mean?
scores are spread more, increase in SD, more variation.
More overlap.
Variation spreads even more. Huge degree of overlap between the 2 groups meaning there is less likely to be a significant difference between them.
What variability is this?
High!!
How would you calculate the t value?
t = difference in means/ variability in means
As the difference between ___________________ increases, the t value increases and is more likely to be significant.
2 means
As variability in means ____ the t value decreases and is less likely to be significant.
increases
aka. we want low variability for significant result!
A ____ t value is more likely to be significant.
larger
What hypothesis is this?
There is no difference between the means of your two groups.
Null Hypothesis
What hypothesis is this?
There is a significant difference between the means of your two groups.
Alternative Hypothesis
What hypothesis can take 2 different forms?
AH
1 tailed - states direction
and 2 tailed- doesn’t state a particular direction.
95% CI for girls’ scores = 2.44, 3.84.
What does this mean?
95% of the time the girls’ scores will be within this range.
If the CIs for the 2 groups overlap what can we infer?
That results aren’t going to be significant!
If the CIs for the 2 groups DONT overlap what can we infer?
That there is a difference between the groups and it’s likely to be significant.
How do we write the t test results?
t (df) = t value, p = .xxx (one/two tailed) (effect size of sig)
What test assesses homogeneity of variance? (e.g. do the 2 groups have similar levels of variation?)
Levene’s Test of Equality of Variances :)
What are we hoping for the Levene’s Test of Equality of Variances?
That it’s a non-significant result!
e.g. > .05
What is a small effect size (cohen)?
0.1
What is a medium effect size (cohen)?
0.3
What is a large effect size (cohen)?
0.5
How do we calculate r (the effect size)?
√t2/ (t2 + df)
Once you have conducted the independent t-test what do you need to remember to check for?
Homogeneity of variance!
using Levene’s Test.
For parametric data what scale does your data have to be on?
Interval/Ratio
For parametric data how many participants do you need in each group?
at least 10 people in each group!
What is alpha and power set at?
alpha 0.05
power 0.80
If you have a between groups design BUT data isn’t parametric e.g. ordinal data, skewed, what test do you use?
Mann-Whitney U
What is the non parametric version of the independent t-test?
Mann- Whitney U
What is the sign for Mann-Whitney U?
U :)
If it’s a one tailed test what do you have to do to the p?
divide it by 2!
Parametric tests are more statistically powerful than non-parametric tests this means you need ____ people to carry out a parametric test.
LESS
If we have more than 3 independent groups and non-parametric data what test do we use?
Kruskall Wallis
What test is an extension of the mann-whitney U for 3 groups?
Kruskall Wallis Test
What do the Mann- Whitney U and Kruskall Wallis involve?
Ranking the data :)
What is the sign for Kruskall Wallis?
H
What test would you get a result like this from?
H(2) = 11.4, p= .003
Kruskall Wallis
After the Kruskall Wallis what do we have to do?
Follow up tests!
What follow up tests do we use for a Kruskall- Wallis?
Follow up Mann-Whitney U tests.
What do we need to remember when conducting follow up Mann - Whitney U tests?
Do Bonferroni Correction!
What would the Bonferroni Correction be if we had 3 follow up tests?
.05/3 = .017
we can add an effect size to a t-test but not to a ….
Mann-Whitney U test
How do you report the Mann-Whitney U test?
U = mann whitney u, p = .xxx (one-tailed)