variance estimates in known fate analysis

[cervidbio]

I have a telemetry dataset that I am using to estimate survival for several demographic classes and test alternative models under the known fate datatype. I am interested in generating annual survival estimates and the dataset contains data from 4 years of monitoring.

I have some censoring issues stemming from a transmitter performance issue associated with 1 particular shipment. So, to maximize the potential information in the data, I have coded the encounter histories using 2-month intervals. Obtaining annual survival estimates is relatively straighforward as the product of the 6 2-month intervals making up a year. What is unclear to me is how to derive variance estimates for annual survival from the 2-month interval estimates.

[bmcclintock]

I have a telemetry dataset that I am using to estimate survival for several demographic classes and test alternative models under the known fate datatype. I am interested in generating annual survival estimates and the dataset contains data from 4 years of monitoring.

I have some censoring issues stemming from a transmitter performance issue associated with 1 particular shipment. So, to maximize the potential information in the data, I have coded the encounter histories using 2-month intervals. Obtaining annual survival estimates is relatively straighforward as the product of the 6 2-month intervals making up a year. What is unclear to me is how to derive variance estimates for annual survival from the 2-month interval estimates.

The "delta method" may be used to approximate the variances of your annual survival estimates. Appendix B (specifically B.4 for your needs) of "A Gentle Introduction" covers this in some detail, and MARK provides the appropriate variance-covariance matrix under the "Output" tab.

[gwhite]

If you break up your 4-year encounter history into 4 1-year encounter histories, the derived parameter will be the annual survival rate, with a variance and confidence interval. The derived parameter in the known fate data type is the probability of surviving the entire interval, so you just have to be cleaver in how you define your histories.

Gary