Home > Statistics Every Writer Should Know > The Stats Board > Discusssion
Familywise error
Message posted by Kathy on August 2, 2000 at 12:00 AM (ET)
1. What is the meaning of familywise error? Is the concept same as multiple comparison problem?
2. When do we use such term? {i.e. t-test or regression} Do we use the term "multiple comparison problem" when we perform ANOVA and when we perform t-test, we use the term "familywise error"?
3. Can we use "Bonferroni " to solve the familywise error?
READERS RESPOND:
(In chronological order. Most recent at the bottom.)
Re: Familywise error
Message posted by Chad Allen on August 8, 2000 at 12:00 AM (ET)
The way I've always understood it, familywise error rate is what is increased when you do an excessive number of inappropriate two group comparisons. For example, say you did a study to see if one religious group (in this case, Baptist, Protestant, and Catholic) had a higher annual income. If you did several (3 in this case) t-tests to compare each group to every other group, your .05 alpha level would be inflated to about .10 or .15. However, if you do an ANOVA, F lets you compare all three groups while keeping alpha = .05.
Now, if the ANOVA's F is significant, you know there is a significant difference somewhere between the groups. Post hoc comparisons (i.e. multiple or pairwise comparisons)will tell you which two groups are significantly different from each other. Bonferonni's is an appropriate post hoc test.Hope I was of some assistance
Chad Allen
Your $5 contribution helps cover part the $500 annual cost of keeping this site online.