c hat, c hat, c hat

[Guest]

Dear all

I have to select the c hat, which I want to use for the adjustment.
Short summary – data set contains two groups (males/females).
The model phi (t) p (t) was tested.
The bootstrap procedure (100 simulations – p level 0.2) gives me 1.02 (deviance approach) and 1.07 (c hat approach).

RELEASE rejects the model with 0.0382. Therefore the c hat is somewhat larger with 1.235. Although RELEASE states that it has not enough data for almost all single tests of TEST3 (SM and SR). The p levels of TEST3 are quiet high (around 0.7) which I doubt due to the data quantity problem. So I do not know if the overall p level (TEST2+ TEST3) is biased due to that.
For the males TEST2 was non significant with enough data while it seems that the females have "a heterogeneity problem" since TEST2 was significant.

U CARE confirmed the latter statement. All four tests for the males were non significant while TEST2.CT for the females was significant – leading to the immediate trap dependence model for that gender. Due to this single test the overall test statistic was significant with 0.0139 and a c hat of 1.291. U CARE does not comment the data quantity problem.

So I have started including trap dependence in my model – but which c hat should I use for the adjustment.

Many thanks
Christian