I’m trying to compare some morphetrics between male and female fish. When I do a t-test some of the measurements show significant difference However when I make an ANOVA on 4 subgroups ( Immature male, mature male, immature female, mature female) no significant difference is found between the corresponding groups ( immature males-females and mature males-females). Is that normal?
You are comparing different models that use different data. You shouldn't expect to get the same results. In some crytal-clear cases the chosen model won't matter much for the conclusions you arrive at, but if the effects are not big (compared to the nose), the model becomes relevant.
Btw, why don't you use a two-factorial model (one factor is "sex" [m/f] and the other is "age" [immature/mature]). If you think that mature animals may show different sex-effects than immature, then you should also include the interaction of "sex" and "age".
It sounds as if you have a couple of variables for which you want to analyze sex-effects (you are screening which variables show a sex effect). If so, then take care to correct for mutipliple testing. Otherwise you will overestimate the confidence in your selection!
You shouldn't expect the same results. In the first case that you used t-test, you have two independent groups while in the second case you have independent four groups.
I think there is problem with how you specified your four groups, which are actually the components of a 2x2 design -- mature versus immature and male versus female. You should re-run the ANOVA with those specifications and examine whether you have two main effects (maturity and gender) or an interaction effect (size varies by a combination of maturity and gender).
I guess that your treatments were,
T1 (Immature males): y11, y21,…,yn1
T2 (Mature males): y12, y22,…,yn2
T3 (Immature females): y13, y23,…,yn3
T4 (Mature females): y14, y24,…,yn4
Therefore, I understand that t-test were made with T11=T1+T2 and T22=T3+T4 and your result indicated significant difference between T11 and T22.
Then, when you performed ANOVA with T1 to T4 your multiple comparison test (Tukey?) did not show significance difference between T1 vs T3 and T2 vs T4.
This can be explained by the fact that the difference found in t-test was due to differences between T1 vs T4 or T2 vs T3. I know that these differences may have no interest, but this can be the reason of your results. You should test the Tukey comparison T1 vs T2 and T3 vs T4.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Modeling
- Analyses of Variance
- ANOVA