c-hat adjustments

[Paul de Tores]

Dear forum members,
I’m seeking some advice with the Known Fate model when making adjustments to c-hat. One of the options in the Preferences settings makes it possible to “add 1 to K when computing QAICc with c-hat >1”. When I have used this option and then subsequently go back to compare models without an adjustment to c-hat, the AICc value does not revert to the original value – the value when there was no adjustment to c-hat. In fact, if I try incrementally increasing c-hat and then revert to the unadjusted value, there seems to be a progressive and irreversible increase in the AICc value (i.e. an increase in the original unadjusted AICc value).

If I retrieve the original model structure and re-run the model, the AICc result is the same as the original, but this seems to be a labour intensive process and not very practical if the set of candidate models is reasonably large.

Can anyone advise on this?

[gwhite]

Paul:
The option you write about is not something that can be turned on and off. It is meant to only be used once, at the very start of the MARK analysis. Generally, I don't use this option.
I also wonder where you are getting c-hat for known fate data, given that the saturated model is a useful model for estimation with the known fate data type. You must be assuming some set of categories have the same survival in order to estimate c.
Gary