chat>1, error return not zero

questions concerning analysis/theory using software from the Patuxent Software Archive

chat>1, error return not zero

Postby storrence » Fri Dec 16, 2005 1:24 pm

Hello.

I am using MARK and I am using rdsurviv to estimate overdispersion on about 24 study sites. For some of these sites, chat < 0 (0.003, 0.69, 0.3908). I also have one site where chat=0.9482.

How close to 1 does the chat estimate have to be to be used? For the value of 0.003, I realize I probably don't have enough data for this model, but what about for the other 2 values?

Do I need to adjust my models in MARK with the chat value of 0.94?

I am also sometimes getting an error-return value of 130. Does my chat value correct for this problem?

Thanks.
storrence
 
Posts: 4
Joined: Wed Mar 30, 2005 12:55 pm

Re: chat>1, error return not zero

Postby cooch » Fri Dec 16, 2005 1:53 pm

storrence wrote:How close to 1 does the chat estimate have to be to be used? For the value of 0.003, I realize I probably don't have enough data for this model, but what about for the other 2 values?

Do I need to adjust my models in MARK with the chat value of 0.94?

I am also sometimes getting an error-return value of 130. Does my chat value correct for this problem?



I suspect Jim Hines (who created RD-SURVIV) will chime in, but to address some of the generic questions....

1. c-hat is so called because its an estimate. That means that there is some error variance we could (in theory) assign to the estimate. You can see a good example of this in the MARK book - chapter 5 (GOF chapter), section 5.8. So, even if the true value of c were (say) 1, c-hat could vary a fair bit above or below this value.

2. if c-hat is >1, might indicate overdipersion in the data. c-hat<1 makes less 'biological sense', so the general recommendation (also noted in Chapter 5), is to set c-ht to 1 if the estimated c-hat is <1.
cooch
 
Posts: 1461
Joined: Thu May 15, 2003 4:11 pm
Location: Cornell University


Return to analysis help

Who is online

Users browsing this forum: No registered users and 1 guest