Estimates of g0 by occasion (g0~t)

[MD_Clements]

We are estimating densities of small mammals (Peromyscus and Microtus) using trapping grids (50-100 traps). We trap once a year in the summer for 1-2 weeks (5-10 occasions) . We have found a very significant effect of occasion on probability of capture (g0~t) for Peromyscus.

We would like to see the trend in the estimates of g0 for each occasion and have tried to use predict() to extract them, but receive an error:

Code: Select all

x<-predict(Pero08fits$secr3,savenew = TRUE)
> x
       link   estimate SE.estimate         lcl         ucl
D       log 67.9346602   8.4750784 53.24892714 86.67062986
g0    logit  0.0601937   0.0130088  0.03921036  0.09133873
sigma   log 12.5824882   0.8421017 11.03728465 14.34401800
> str(x)
'data.frame':   3 obs. of  5 variables:
 $ link       : Factor w/ 2 levels "log","logit": 1 2 1
 $ estimate   : num  67.9347 0.0602 12.5825
 $ SE.estimate: num  8.475 0.013 0.842
 $ lcl        : num  53.2489 0.0392 11.0373
 $ ucl        : num  86.6706 0.0913 14.344
 - attr(*, "newdata")='data.frame':   1 obs. of  2 variables:
  ..$ session: Factor w/ 1 level "Bumpkin 2008 Grid C": 1
  ..$ t      : Factor w/ 9 levels "1","2","3","4",..: 1
>predict(Pero08fits$secr3,newdata = data.frame(t=0:9))
Error in mat %*% beta[indx] : non-conformable arguments


Any assistance would be greatly appreciated.

M

[murray.efford]

Before I dig deeper - I don't think t=0 is a valid factor level: don't your times run 1:9? Use T if you actually want a linear trend on the logit scale. Have you considered a learned response?
Murray

[murray.efford]

The critical point is that you need to make a factor from the integer occasion numbers

Code: Select all

newdata = data.frame(t = factor(1:9))

etc.
Murray