GOF for AS model not running

[monicaarso]

Hi,
I am running GOF for Multistate in U-CARE, based on 190 individual capture histories, with 3 sex states (Unknown, Female, Male). The input MARK type file looks like this, where numbers 2,3,4 are the 3 states:

Code: Select all

22022222220222222444444 1;
00222020202220333333333 1;
22020202000020200000000 1;
20000000000000000000000 1;


When I run the Sum of multisite tests in the Goodness-of-Fit for Multistate menu, only Tests 3G, M and FOG for JMV model are run, but not the GOF for the AS model as described in Chapter 8 of MARK book. See output for tests:

Code: Select all

GLOBAL MULTISITE TEST, group =1
------------- Test 3G (sta, pval, df) ----------------------------------
72.289   0.567  75.000
------------- Test M (sta, pval, df) ----------------------------------
35.385   0.000   9.000
------------- Goodness-of fit Test for the JMV Model (sta, pval, df) ----------------------------------
107.674   0.042  84.000
*******************************************

Any ideas why GOF for AS model doesn't run?
Many thanks in advance!
Monica

[Guillaume Souchay]

Hi Monica,
I don't know if you still need help for your analysis.

Any ideas why GOF for AS model doesn't run?

Maybe you miss something if you did not read the U-CARE manual. In U-CARE, the multistate test is based on the JollyMove model which is more general than the AS model. So, the result provided by U-CARE are the good one.

Guillaume

[monicaarso]

Dear Guillaume,
thanks so much for your reply. I got confused by what the MARK book explains on GOF for Multistate and the example they give (which shows both AS model and JMV model).
merci!

[Guillaume Souchay]

Dear Monica,
Glad it help you.

Do not hesitate to ask other questions if you need !

Guillaume

[cooch]

Hi Monica,
I don't know if you still need help for your analysis.

Any ideas why GOF for AS model doesn't run?

Maybe you miss something if you did not read the U-CARE manual. In U-CARE, the multistate test is based on the JollyMove model which is more general than the AS model. S...

Which is precisely what the MARK book says:

to assess the fit of the fully time-dependent JMV model to the data. In many cases, the JMV model is unlikely to show significantly greater t to the data than the AS model (since the dependence of the capture probability at time (i+1) to depend on both (i) and (i+1) is unlikely to be observed in practice very often). As such, the GOF test for the JMV model may by a generally valid test of t for the AS model as well...

and, the example as described actually refers only to the JMV result.