Closure test differences between SECR and CAPTURE ?

[Lucas Luche]

Hi, anyone know the details about pratical differences between closure test in SECR and in conventional . 

I'm comparing the results of both in multiple periods and in the SECR ones the Z values seens more likely to stay negative (non-closed) ... with the capture history, of course. 




Big thanks 

[Murray Efford]

Your question is vague.

[Lucas Luche]

Sorry... 

It's that I'm getting different results for Closure Test from SECR package and CAPTURE, using the same capture history.
I'm observing that the SECR ones tend to more likely consider a open population ( negative Z values), but I can find any theorical justifications for that and I wanna know if anyone gets it or knows where I can find it 


The SECR standart closure test is based in Stanley & Burnham (1999) , while CAPTURE's is based on Otis (1978), right ? But Ia justification for those diiferences in the original documents 

Thank you very much 

[Murray Efford]

The R package is 'secr' not SECR. I assume you refer to the function 'closure.test'. As you would discover from reading the help page, closure.test defaults to exactly the same test as CAPTURE (i.e. the Otis et al 1978 test). You do not give an example.

For the default dataset in DENSITY, CAPTURE gives
Overall test results --
z-value                                 -1.408
Probability of a smaller value         0.07949

For the same dataset in 'secr'

closure.test(captdata)
statistic         p 
 -1.40849   0.07949 

I don't know what you are talking about. The Stanley & Burnham test may be selected as an option in closure.test. None of the above should be taken to indicate I think closure tests are necessary or informative.

Murray

[Eric H]

Hi Lucas,

You might get different results between closure.test ('secr' package) and CAPTURE if your SECR data include additional recaptures (e.g. at different traps on the same occasion - not included in CR data). 

Note that each of the closure tests have limitations.  I think it's the test of Stanley and Burnham that tends to reject if there is heterogeneity in capture probabilities among individuals - larger samples may increase the chance that this test will reject if there is heterogeneity (which there usually is).  The other test has low power to detect violation, but tends to reject in the presence of behavioural responses to detection... This is an incomplete description - best to look at the original references.

All the best,

Eric