seasonal covariates in multi-season occupancy
Posted: Wed Oct 31, 2012 6:28 pmGreetings,
I am running a multi-season site occupancy analysis and am having problems coding seasonal covariates for each Psi in the RDOccupPE/PG models. I have searched the Appendix/phidot posts and found some similar examples, but nothing close enough to my specific problems to help.
Background – I am examining occupancy of nests for two seasons in relation to a variety of habitat and disturbance covariates. I am less interested in factors that influence p and used a “removal design” to survey for nests, resulting in missing observations in the ch string after initial detection. I would like to estimate Psi for each year because I have seasonal covariates that need to be coded separately. I would like to modify the Design Matrix in the following way as I have done in MARK where each covariate is specific to the year.
Parameter B1 B2 B3 B4
Psi-1 1 covar1_2010 0 0
Psi-2 1 covar1_2011 0 0
eps or gam 0 0 1 0
p 0 0 0 1
Unfortunately, I cannot figure out how to code different covariates for each Psi. As far as I can tell, I need to get RMark to recognize the suffix in the covariate names (covar1_2010/2011), as in the td function in section C.16 of App_C (p. C-78), and then treat the covariate as time-varying. However, this function doesn’t work for my data because of the missing values in the ch string. Also, I have tried to experiment with the time+covar and covar:time models below, but found the following results to not achieve my desired DM either.
formula=~time+covar
1 0 covar 0 0
1 1 covar 0 0
formula=~covar:time
1 covar 0 0 0
1 0 covar 0 0
I have attached a snippet of my code below. Any help would be greatly appreciated.
Thanks,
Patrick
I am running a multi-season site occupancy analysis and am having problems coding seasonal covariates for each Psi in the RDOccupPE/PG models. I have searched the Appendix/phidot posts and found some similar examples, but nothing close enough to my specific problems to help.
Background – I am examining occupancy of nests for two seasons in relation to a variety of habitat and disturbance covariates. I am less interested in factors that influence p and used a “removal design” to survey for nests, resulting in missing observations in the ch string after initial detection. I would like to estimate Psi for each year because I have seasonal covariates that need to be coded separately. I would like to modify the Design Matrix in the following way as I have done in MARK where each covariate is specific to the year.
Parameter B1 B2 B3 B4
Psi-1 1 covar1_2010 0 0
Psi-2 1 covar1_2011 0 0
eps or gam 0 0 1 0
p 0 0 0 1
Unfortunately, I cannot figure out how to code different covariates for each Psi. As far as I can tell, I need to get RMark to recognize the suffix in the covariate names (covar1_2010/2011), as in the td function in section C.16 of App_C (p. C-78), and then treat the covariate as time-varying. However, this function doesn’t work for my data because of the missing values in the ch string. Also, I have tried to experiment with the time+covar and covar:time models below, but found the following results to not achieve my desired DM either.
formula=~time+covar
1 0 covar 0 0
1 1 covar 0 0
formula=~covar:time
1 covar 0 0 0
1 0 covar 0 0
I have attached a snippet of my code below. Any help would be greatly appreciated.
Thanks,
Patrick
- Code: Select all
time.intervals=c(0,1,0)
RD_process=process.data(RD.test, model="RDOccupPE",
time.intervals=time.intervals, begin.time=2010)
RD_ddl=make.design.data(RD_process)
# Creating list of potential predictor variables for Psi
Psi.dot=list(formula=~1)
Psi.time=list(formula=~time)
Psi.time.plus.covar=list(formula=~time+covar)
Psi.covar.by.time=list(formula=~covar:time)
# Creating list of potential predictor variables for p
p.dot=list(formula=~1)
p.time=list(formula=~time) #by secondary
p.time.plus.session=list(formula=~time+session) #by secondary
p.session.by.time=list(formula=~time:session) #by secondary
p.session=list(formula=~session)
# Creating list of potential predictor variables for Gamma
gam.dot=list(formula=~1)
epsilon.dot=list(formula=~1)