Interpretation of significant 3G.Sm test

questions concerning analysis/theory using programs M-SURGE, E-SURGE and U-CARE

Interpretation of significant 3G.Sm test

Postby Whidden » Mon Mar 16, 2015 8:49 pm

Hello,

I'm running GOF tests in U=CARE for a juvenile survival and movement analysis in MARK on seabirds.
Summed results are significant (χ2 = 500.04, df = 217, p=0.000). Individual tests show significance for:

WBWA (x2 = 154.62, df = 26, p = 0.000)
3G.Sm (x2 = 231.97, df = 115, p = 0.000)
M.ITEC (x2 = 58.97, df = 23, p = 0.000)

but not for:

3G.SR (x2 = 14.985, df = 21, p = 0.82)
M.LTEC (x2 = 39.48, df = 32, p = 0.17)

So, I need to account for memory (still researching how to do this) and trap-dependence (through an individual covariate) in the MARK models. However, I'm struggling with the interpretation of a significant 3G.Sm test in the presence of a non-signficant 3G.SR test. I know from the U-CARE manual that 3G.SR tests for the presence of transients, but I am having some trouble with interpretation of what 3G.Sm is actually testing for, and thus, how to account for it in the models.
Any insight would be greatly appreciated.

Thanks,
Erin
Whidden
 
Posts: 1
Joined: Tue Jun 05, 2012 7:09 pm

Re: Interpretation of significant 3G.Sm test

Postby Guillaume Souchay » Tue Mar 24, 2015 10:43 am

Hi Erin,

For goodness-of-fit of multistate models, to sum up, I would say that the WBWA test deals with "memory" in your model (e.g. individuals stay in some states more than expected, see also Cole et al. 2014 Ecology and Evolution 4:2124-2133). The 3G.SR test is for presence of transient in your dataset and test 3G.Sm is "just" a complement test. Finally, tests M.ITEC and M.LTEC deal with trap-dependence. The M.ITEC test deals with immediate trap-dependence (difference in encounter rate at time t+1 between animals in same state at time t) while the M.LTEC test is more about long-term trap-dependence (somewhat linked to some memory effect in the data - maybe Rémi could give his opinion).

Based on the result from the GoF test you ran, there is some evidence for memory effect in your dataset. One way to deal with memory is proposed by Rouan et al (2009 Journal of Agricultural, Biological, and Environmental Statistics 14:338-355). However, memory model are not possible with MARK.
Instead of dealing with transition from A to B (with all combinations), you are now taking into account state at time t-1. Contact me also if you need more explanation/help.
Beside of the memory effect, you can just compute a c-hat without the result of the WBWA test.

Hope this can help.

Cheers,
Guillaume
Guillaume Souchay
 
Posts: 64
Joined: Wed Feb 27, 2008 5:20 am
Location: Nantes, FRANCE


Return to analysis help

Who is online

Users browsing this forum: No registered users and 13 guests

cron