Family-Wise Error & Assumptions Flashcards
What is a family-wise error?
An accumulated Type I error (from a whole set of tests).
What is an error on individual tests called?
Per comparison error.
What is the family-wise error rate?
The probability of making at least one Type I error from 𝑚 comparisons.
If our comparisons are planned, we don’t need to correct for FWE. Why is this?
Because the probability of making an error is modified by our prior expectation that the null hypothesis is false (an informal application of Bayes’ theorem).
If we are performing post-hoc comparisons, what do we need to do to account for FWE?
Adjust our error rates.
What is the most conservative FWE correction? What is the risk with this?
Bonferroni.
The risk is that is ensures a very high false negative rate.
What is the Bonferroni correction often performed on?
t-tests.
What does the Scheffé Test do?
Adjusts the critical F-value using the number of treatment levels.
What does the Tukey Honestly Significant Difference (HSD) test use to limit Type I error? What is this used for?
The studentised range distribution Q.
This is used to estimate the probability of observing a range in DV values for a sample.
Post-hoc tests reduce the chance of _______ by greatly increasing the chance of _______.
Type I errors, Type II errors.
What are the assumptions of ANOVA?
- Independence (treatment levels are randomly assigned)
- Homogeneity of variance (only the means differ)
- Normality
A common test for homogeneity of variance in a between-group design is _________.
Bartlett test.
A common test for homogeneity of variance in a mixed or within-group design is _________.
Box’s M.
What are 3 common tests for normality? What are they very sensitive with?
- Skewness
- Kolmogorov-Smirnov
- Shapiro-Wilk
These are very sensitive with large samples.
A perfectly normal distribution has a skewness of _____.
Zero.
