GOF tests with missing values

[tarwater]

I am using recapture only data with missing values to estimate survival. GOF tests cannot be run using encounter histories with missing values. Running a GOF test on a smaller subset of my data (n=134) results in a c-hat close to 1. Survival estimates from the subset of data are similar to the survival estimates based on the complete data set (n=198). Is it valid to use the c-hat from the subset of data on my larger data set? Any suggestions?

[cooch]

I am using recapture only data with missing values to estimate survival. GOF tests cannot be run using encounter histories with missing values. Running a GOF test on a smaller subset of my data (n=134) results in a c-hat close to 1. Survival estimates from the subset of data are similar to the survival estimates based on the complete data set (n=198). Is it valid to use the c-hat from the subset of data on my larger data set? Any suggestions?

Not sure what you mean by 'missing values'. Do you mean occasions where you didn't sample, or individuals not detected on some occasions? If the latter, then this is obviously not an issue, since the estimation is designed to account for non-detections. If the former, this would lead to data sparseness, which may or may not be a large issue depending on how many occasions were 'missed'.

[tarwater]

I mean occasions where I did not sample, so I have "."s in the encounter history (6 intervals). I have 134 individuals with no "."s and 64 with missing values. In 41 out of the 64 individuals, there is only one missing interval in the encounter history. The rest have 2 missing values. If I only include the individuals with one missing value would that reduce any potential biases?