[R] Tukey, planned contrasts or t-test without ANOVA? What is correct?

[Dieter Menne]

J S <yulya258 <at> hotmail.com> writes:

> I compare mean monthly body temperature between two age classes of turtles
overwintering underground. 
> 
> lm(body_tem ~ Month*Year*Age_Class)
> TukeyHSD(aov(body_tem ~ Month*Year*Age_Class, a))
> 
> The Tukey HSD as well as the planned contrasts method showed significant
differences between the two age
> classes, but insignificant differences between the two age classes at the same
levels of months. 
> 
> In the opposite, using a t-test for comparison of independent means (i.e.
without using the ANOVA) it
> demonstrated insignificant differences between the two age classes. What
result is correct?

If the more detailed breakdown shows an effect, use it, because it tells you 
that grouping was over different classes.

Dieter