Delta-method variance-covariance matrix

[Fabienne Sutter]

Hi
I calculated the standard errors of total survival of different tadpole communities following the delta method (mark book, pages B2-B15). That went pretty well, thanks. The variance-covariance-matrix necessary for the calculation of the standard errors was generated in program MARK and imported into “Scilab” (Matlab clone). In cases where I used model-averaging for the calculation of phi there’s no possibility to generate a variance-covariance-matrix in program MARK. In order to get estimates of standard errors of averaged phi’s I tried to generate the variance-covariance-matrix in “Scilab”. Unfortunately I don’t get the right Matrix – probably because I wrongly arrange the phi’s in the vector?

Here’s the Scilab-code I entered for the calculation of the variance-covariance-matrix (it should work in MATLAB too). On request I send the “Scilab-code” for the calculation of standard errors of phi.


// Code beginning
phi1=0.9
phi2=0.8
phi3=0.85

// making vector phi
phi=[phi1 phi2 phi3]

// calculation of variance-covariance-matrix “v” with command “mvvacov
v=mvvacov(phi)
// Code end


what I should get, but I was unable to calculate:
v=[0.02 0.001 0.001;
0.001 0.03 0.005;
0.001 0.001 0.04];

I would appreciate any hint to this problem.
Many thanks and best regards
Fabienne

[Fabienne Sutter]

I've got the answer to my question...

You can get the variance-covariance matrix of model averaged parameters
from MARK by clicking on the check box in the lower left corner of the
model averaging parameter selection window...

Fabienne