GOF in UCARE - do the directional tests override the chi2?

[cathrumm]

HI
I'm comparing my Goodness of Fit result from RELEASE (through MARK) to one i got in UCare, and Im unclear about what Ucare is saying as mixture of sig and non-sig results. In RELEASE The CJS model is non-significant. In UCARE:The quadatic chi2 results suggest non-sig, but the directional tests suggest significant result for trap response and borderline for transience (one approach says NSig, but LOR approach says Sig).

If i add up the separate components from the result of UCare to test :
phi a2*t, pt ( ie 3.Sm+2.Ct+2.Cl) = NonSig
then phi a2*t, pt versus phi t, pt (3.Sm+2.Ct+2.C +2.SR) = NonSig
phi t, pm*t (3.SR + 3.Sm +2.Cl) = non sig
phi t, pm*t versus phi t pt (3.SR + 3.Sm +2.Cl+2.Ct) = Non sig

BUT the directional tests:
transience - the first approach is non-sig, but the 2nd ie LOR = SIG
trap response - both approach give Sig results.

Is this because directional are more sensitive? Hence, do i take a Sig result in the directional test to mean apply that model? In which case choose trap response over the tranisent model?

Thanks
Cath Rummery