Difference in deviance and number of capture histories

This may have been posted before but a contact with someone offlist made me realize that a reminder may be useful. The deviance value put in the marklist model table is the residual deviance. If you are comparing models with and without covariates the deviance values will be quite different because the null deviance with covariates is set to 0 by MARK. You can use a switch in model.table to use -2lnl instead of deviance.
In trying to work this out they looked at the output files from MARK and saw that the number of capture histories differed between the 2 runs. Beginning in v2.1.4 the code accumulates data records that are the same and uses the freq field for the number of records with that value. This is not necessary for models without individual covariates because mark.exe does the same accumulation before it fits the model. But it doesn't do that accumulation if the data have individual covariates. Sometimes individual covariates only have small number of possible values so it is possible to accumulate. RMark will accumulate identical records in all cases unless you set accumulate=FALSE. This can speed up run times when there are individual covariates. Even though the number of data records differ the sum of the frequencies will be the same.
--jeff
In trying to work this out they looked at the output files from MARK and saw that the number of capture histories differed between the 2 runs. Beginning in v2.1.4 the code accumulates data records that are the same and uses the freq field for the number of records with that value. This is not necessary for models without individual covariates because mark.exe does the same accumulation before it fits the model. But it doesn't do that accumulation if the data have individual covariates. Sometimes individual covariates only have small number of possible values so it is possible to accumulate. RMark will accumulate identical records in all cases unless you set accumulate=FALSE. This can speed up run times when there are individual covariates. Even though the number of data records differ the sum of the frequencies will be the same.
--jeff