I have a variable that I can divide to 3 groups: A, B, C. I want to check if the difference between the groups is significant. Should I use F-STATISTIC or maybe T-STATISTIC of each group vs. the other 2 groups together?
$\begingroup$
$\endgroup$
Add a comment
|
1 Answer
$\begingroup$
$\endgroup$
1
I would use ANOVA or Kruskall-Wallis Test, depending on the normality of data and other assumptions.
I guess that your null hypothesis is as follows (correct me if I'm wrong):
$$ \mu_A = \mu_B = \mu_C $$
Since you have more than two groups, it will be better to use the mentioned tests than $t$ test. For example, with ANOVA (which uses the $F$ statistics) you perform one test to check if any group is significantly different than the other, but you would need 3 $t$ tests to do the same: A~B, A~C, and B~C.
-
1$\begingroup$ Thanks a lot for the fast reply, I thought so but wanted to be sure :-) $\endgroup$ Commented Apr 18, 2016 at 15:57