Why I get significant results with t-test when I compare two samples individually, but when I do two ways ANOVA followed by post hoc test I can not see the difference between that specific individuals?
Hi Sara,
This non-significance is due to different numerical-computaional formulas for deriving the error estimates (which are more restrictive in posthoc-pairwise comparison tests).
Hope this helps,
Matt
If the null hypothesis is true, then setting alpha=0.05 to decide if you have a significant difference is saying that you will mistakenly declare a significant difference when none exists (the null hypothesis is true, so this is a mistake). In this scenario, the more tests I make the more likely it is that I will find at least one significant difference. The t-test does not correct for this problem. The other tests do correct the problem but there is a cost. The simplest correction is the Bonferroni correction where alpha is divided by the number of pairwise comparisons. However, this test is often too restrictive (we fail to reject the null hypothesis more often than we want).
anova is a global test of reationship of v1 and v2 but neither tells you which categories are related nor the direction or strength of the relationship
- Badges
- Science topic