insufficient data for GOF in U-CARE

[ben]

I am trying to perform GOF tests for single state models (live encounters). I did these tests separately for adult females and for adult males.

For adult females, the simple CJS model is acceptable : there is no transient effect (two-sided test P =0.5431,one-sided test P=0.7284) and no trap-dependence effect ( two-sided test P=0.36575)

However for adult males I obtained the following results :
Global TEST, number of groups =1
df =6
Quadratic Chi2 =6.1232
->P-level=0.40953

N(0,1) statistic for transient(>0) =0.7667
->P-level, two-sided test =0.44326
->P-level, one-sided test for transience =0.22163


N(0,1) signed statistic for trap-dependence =NaN
->P-level, two-sided test =NaN

Does it mean that it is not possible to test the trap-dependence from my matrix of individuals’ histories ? i.e. insufficient data to make the contingency tables ?

Does the fact that it’s not even possible to test the trap-dependence, indicates that one can neglect trap-dependence in modeling the data?
In other words, is a CJS model “acceptable” in such a case?

[cooch]

I am trying to perform GOF tests for single state models (live encounters). I did these tests separately for adult females and for adult males.

For adult females, the simple CJS model is acceptable : there is no transient effect (two-sided test P =0.5431,one-sided test P=0.7284) and no trap-dependence effect ( two-sided test P=0.36575)

However for adult males I obtained the following results :
Global TEST, number of groups =1
df =6
Quadratic Chi2 =6.1232
->P-level=0.40953

N(0,1) statistic for transient(>0) =0.7667
->P-level, two-sided test =0.44326
->P-level, one-sided test for transience =0.22163


N(0,1) signed statistic for trap-dependence =NaN
->P-level, two-sided test =NaN

Does it mean that it is not possible to test the trap-dependence from my matrix of individuals’ histories ? i.e. insufficient data to make the contingency tables ?

Does the fact that it’s not even possible to test the trap-dependence, indicates that one can neglect trap-dependence in modeling the data?
In other words, is a CJS model “acceptable” in such a case?

When in doubt, fit TSM models (2 classes) as your general model. This is the most conservative approach.