I was trying to run a Huggins Full Heterogeneity (HugFullHet) model with group effect. However, these 2 "groups" were actually "strata" among the population. To be more specific, I was trying to estimate abundance of a species from 2 grids in my study area. I looked over "multi-strata" model but I wasn't interested in their transition between different "states". Not many movements or transitions occurred during sampling so I don't think multi-strata is the type of model I need.
As I tried to apply HugFullHet to my data, it seemed that Huggins required sampling occasions to be consistent among groups. My data from the two grids had different sampling occasions and the model ran with error.
- Code: Select all
ch freq grid
3004 00010000000000000 1 N
3005 00010000000000000 1 N
503 10100000000000000 1 N
522 01000000000000000 1 N
536 00100101000000000 1 N
537 00100100000000000 1 N
539 00110010101000000 1 N
540 00100010000100000 1 N
541 00100010000000000 1 N
543 00100000000100000 1 N
545 00000000100000000 1 N
546 00010001000000000 1 N
547 00010000000000000 1 N
548 00000010000000000 1 N
554 00000000001000101 1 N
555 00000100100000000 1 N
558 00000010000000000 1 N
560 00000010000000000 1 N
570 00000000010000001 1 N
526 110100000100100010 1 S
527 110000000000000000 1 S
528 100001000000000000 1 S
529 010000000000000000 1 S
530 100000000000000000 1 S
531 110000000000000010 1 S
532 001000100010001000 1 S
534 001101110000100001 1 S
542 001000001010000010 1 S
544 000100000000000000 1 S
5451 000100000000100000 1 S
562 000000010000000000 1 S
566 000000000100000000 1 S
568 000000000000000001 1 S
572 000000000000010000 1 S
573 000000000000001000 1 S
577 000000000000010000 1 S
> mark(data,model="HugFullHet", groups="grid", allgroups=TRUE, model.parameters=list(formula=~grid,share=TRUE))
Error in process.data(data, begin.time = begin.time, model = model, mixtures = mixtures, :
Capture history length is not constant. ch must be a character string
row numbers with incorrect ch length 526,527,528,529,530,531,532,534,542,544,5451,562,566,568,572,573,577
Of course I could run 2 grids separately (which I already did) but it didn't fit my research question. I sampled the study area (with multiple grids) almost concurrently and I think it's more reasonable to estimate N for each grid conditional on the entire population of the study area. Is there away to fix this? If it cannot be done in RMark do I need to go with a BUGS & Bayesian framework instead?
Any comments or suggestions are appreciated. Thanks.
Lamu