Wrong number of encounter histories - output files

[C. Le Coeur]

Dear all,

I’m running some Pollock’s and Huggin’s Robust Design models from two datasets assuming 70 and 712 encounter histories respectively. I’m using both RMark and Mark programs.
Whatever the models are (and the software used), there is a wrong number of encounter histories highlighted in output file (hist=).
The first 2 lines from the output are shown below :

Dataset of 712 Encounter histories (Huggins RD, RMark)

INPUT --- proc chmatrix occasions= 152 groups= 16 etype= RDHuggins
INPUT --- Threads= -1 ICMeans NoHist hist= 549 ;
Number of unique encounter histories read was 458.

Dataset of 70 Encounter histories (Pollock’s RD, MARK program) :

INPUT --- proc chmatrix occasions=25 groups=2 etype=Robust NoHist
INPUT --- hist=300;

Models run ‘correctly’ as far as this issue impacted on. Number of groups and occasions were defined well.

This issue seems to be a common one, as it has been found for others analyses (multi-states models) on different datasets (from colleague's results).

Why the number of encounter histories is incorrect ? What is the impact on modeling ?
What should I do to avoid this ?

Thank you for your help !

C. Le Coeur

[cooch]

Number of unique encounter histories read was 458.

Key word is 'unique'. MARK summarizes individuals that have the same encounter history, such that the number of unique encounter histories is generally < total individuals. I suspect this is what you're seeing.

[C. Le Coeur]

I agree with the number of unique encounter histories (458) < total individuals (712) but I don't understand why I've got a different number (549 instead of 712) for the 1st example and a higher number (300 instead of 70 !) for the second one ?

Total individuals = 712
INPUT --- Threads= -1 ICMeans NoHist hist= 549 ;
Number of unique encounter histories read was 458.

Total individuals = 70
INPUT --- NoHist hist=300;

[jlaake]

Are you using groups? Send me the inp and out files offlist if you can't work it out.

--jeff (jeff.laake@noaa.gov)

[C. Le Coeur]

Thanks a lot for your help. Now, I have a good understanding of encounter histories reported.

With RMark, 3 different numbers of encounter histories are recorded:
- Number of individuals (total of frequencies)
- Hist : number of histories in the input file, i.e sum of ‘unique’ (=single) encounter histories per group
- Number of unique encounter histories read
Whereas Mark interface provides a max to read (hist) and the number of unique encounter histories.

Assuming the same dataset and analysis on MARK interface and RMark, here it is the first 2 lines from the output.

MARK interface:
70 Encounter histories (Pollock’s RD):
INPUT --- proc chmatrix occasions=25 groups=2 etype=Robust NoHist
INPUT --- hist=300;
Number of unique encounter histories read was 62.

=> 70 is the number of individuals.
=> 300 is the maximum to read.
=> 62 is the number of unique histories.

RMark:
70 Encounter histories (Pollock’s RD):
INPUT --- proc chmatrix occasions= 25 groups= 2 etype= Robust
INPUT --- Threads= -1 ICMeans NoHist hist= 66 ;
Number of unique encounter histories read was 62.

=> 70 is the number of individuals (total histories).
=> 66 is the number of histories in the input file, i.e sum of the unique histories per group (in this example, 2groups: male/female). Details about the histories are reported in the input file.
=> 62 is the number of unique histories.

C. Le Coeur