Hi everyone,

Is it always wrong to compare two groups if their means are not significantly different, as measured in a t-test, if they are distinguished on the grounds of a categorical variable (in this case a yes/no variable about demographic information) and it can be shown that the DV in each group correlates significantly with different IVs in each case?

For example, if in a dataset, performance in an exam has a normal distribution, and no significant difference in the mean between males and females can be found in a t-test, but performance for males correlates significantly with hard work and for females correlates significantly with intelligence (so you might conjecture they get the same result by different routes), is this a valid statistical finding, or should everything stop because the t-test is not significant.

Thanks in advance.