Friedman and Kruskal-Wallis Flashcards
What type of tests are Friedman and Kruskal Wallis?
Omnibus tests
What are Omnibus tests?
tests which investigate if there are overall differences between several conditions
BUT doesn’t tell us where difference lies
What does comparing more than 2 conditions lead to?
leads to an increased likelihood of making a type 1 error (false positive)
doing multiple tests means likely to find a difference at least once
What is another name for type 1 error?
familywise error
What do omnibus tests reduce?
the familywise error rate (type 1 error)
What is a type 2 error?
false negative
How do you limit the chances that you’ll commit a type 1 error?
by doing a bonferroni correction
What is a Bonferroni correction?
changing your alpha level to correct for no. of statistical tests we conduct
divide level (.05) by amount of tests you’re going to conduct
e.g 4 tests, .05 / 4 = .0125
What are Post-Hoc tests?
assess the differences between conditions following an omnibus test to find where effect it
ONLY if omnibus test shows a significant result
can be pairwise or planned
When is a Post-hoc test conducted?
after an omnibus test that gains significant result
What is a pairwise comparison?
refer to tests conducted for every pair of conditions
e.g not sure where the diff lies so test everything
What are planned comparisons?
refer to tests you have pre-planned
e.g test a subset
What is the Freidman’s test?
non-parametric
3+ conditions
within subjects design
ordinal or interval / ratio data that’s skewed
What is Friedman’s test sometimes called?
Friedman’s ANOVA
What is Friedman’s test based on?
ranking the data and comparing mean rank of each condition
What is the effect size for Friedman’s test?
Kendall’s W
small = 0.1
medium = 0.5
large = 0.8
What is a Conover test?
a pairwise test
takes the ranks that the freedman test used and analyses them instead
What is the Kruskal-Wallis Test?
Non-paametric
3+ conditions
between subjects design
ordinal or interval/ ratio data that’s skewed
What is the effect size for Kruskal-Wallis test?
Eta squared (n2)
What does Eta Squared tell us?
how much variance in our result is accounted for by our IV (in form of a percentage)