Delta method

[Shawn Craik]

I'm model averaging estimates of nest survival (top 2 models) and was wondering if the derived 95% CI estimates calculated by MARK are the result of the Delta Method? The derived parameter estimate of survival seems right (e.g., product of survival estimates), I just want to make sure the 95% CI estimates are ok.

I'd appreciate any feedback.
Shawn
srcraik@hotmail.com

[cooch]

I'm model averaging estimates of nest survival (top 2 models) and was wondering if the derived 95% CI estimates calculated by MARK are the result of the Delta Method? The derived parameter estimate of survival seems right (e.g., product of survival estimates), I just want to make sure the 95% CI estimates are ok.

I'd appreciate any feedback.
Shawn

No - the Delta method does not feature in the calculation of SE or 95% CI for model averaged parameter estimates. Have a look at section 6.4.1 in Chapter 6 (of the most recent version of the chapter). You should be able to figure out how things work from that.

[bmcclintock]

Shawn,

To clarify, the SEs for the derived parameters of your top 2 models are obtained via the delta method, and the 95% CIs are calculated using these SEs. When you model average over these 2 models, MARK weights the derived parameters and their SEs (that were obtained via the delta method) based on the models' AICc weights. The 95% CIs for these model-averaged estimates are then calculated in the usual way (based on the standard normal distribution) using the model unconditional SEs. As Evan pointed out, the delta method does not enter the equation for model-averaging (although the delta method did play a part in obtaining the SEs of the derived parameters). Bottom line: your CIs from MARK are ok.

Cheers,
Brett McClintock