www.phidot.org

As it states in the documentation, RMark adjusts the parameter counts that MARK reports if they disagree. RMark assumes each parameter is identifiable to avoid under-counting of parameters that end at boundaries (eg p=0 or Phi=1). I chose this default for the argument adjust=TRUE because I have seen cases where MARK under-counts the number of parameters and the unsuspecting user tries to explain variation in parameters that is not supported by the data. My personal philosophy that it is better to err by over-counting parameters but as a user you have complete control over parameter counting.

Any discrepancy between MARK and RMark is reported in the model summary. Below is an example for Phi(t)p(t) with the dipper data.

Code: Select all

> mark(dipper,model.parameters=list(Phi=list(formula=~time),p=list(formula=~time)))

Note: only  11  parameters counted of  12  specified parameters
AICc and parameter count have been adjusted upward

Output summary for CJS model
Name : Phi(~time)p(~time)

Npar :  12  (unadjusted=11)
-2lnL:  656.9502
AICc :  681.7057  (unadjusted=679.58789)


You can use the MARK parameter counts by setting adjust=FALSE in your call to either mark or mark.wrapper as shown below:

Code: Select all

> mark(dipper,model.parameters=list(Phi=list(formula=~time),p=list(formula=~time)),adjust=FALSE)

Output summary for CJS model
Name : Phi(~time)p(~time)

Npar :  11
-2lnL:  656.9502
AICc :  679.5879


In some cases neither parameter count is correct when there is a mix of both parameters at boundaries and non-identifiable parameters as in phi(t)p(t). In those cases you can use the function adjust.parameter.count to adjust the number of estimated parameters and the resulting AIC.

regards --jeff