L3 Flashcards
What ANOVA stand for?
Analysis of variance
What two characteristics can a data set be described by?
- it’s typical value
- how spread out it is (dispersion)
What does ANOVA compare?
Within-group error variance
to
Between-group variance
What would it mean for the variance if both samples were from the same population?
Variance within each group is similar to the variance between groups.
What is the f ratio?
between groups variance
____________________
within-group variance
What does the f ratio test?
The likelihood of two variance estimates being from the same population
What does a significant ANOVA tell us?
That at least one group is different from at least one other group, somewhere in the analysis.
Name two types of follow up analyses.
- Planned comparisons (a priori)
- Post Hoc (a posteriori)
What is the problem with multiple tests/comparisons?
Increases familywise error rate
What are type 1 errors?
Errors involving believing there is a difference when there actually isn’t.
What are type 2 errors?
Errors involving there is no difference when there actually is.
The chance of making a type 1 error is equal to what?
The p value
When you have more than one comparison, what needs to be done
Divide the error rate (0.05) by the number of comparisons to make it stricter.
Lower error rates lead to more _____ criterion for significance.
Strict
How can we get around familywise errors?
- Making the error rate more strict
- Using post hoc tests (e.g. bonferroni)
What level of data is needed for ANOVAs?
- Interval data
- Ratio data
What are the assumptions of non-parametric tests?
- Distribution free.
- Ordinal data (simple treatment of data)
Non parametric tests lead to a greater chance of type ___ errors?
Type 2 errors
What is the Kruskal-Wallis test equivalent to?
Non parametric independent one way ANOVAs.
How many groups does the Kruskal-Wallis test compare?
3 or more
Which test is the Kruskal-Wallis test similar to and why?
Mann-Whitney U because it works on signed ranks.
Once data has been ranked, what does the Kruskal-Wallis test look for?
Systematic differences
What is Friendman’s test a non-parametric equivalent of?
Repeated measures one way ANOVA.
How many conditions does Friedman’s test compare?
3 or more related ones.
