Large (ish) SE's

questions concerning analysis/theory using program MARK

Large (ish) SE's

Postby j.harv3y » Thu Mar 12, 2020 11:44 am

Hi,
I've been running just a CJS and I am getting what I think are quite large standard errors. They're not so massive that I think there might be an issue with identifiability ect. and can follow advice given in the handbook. I was just wondering if anyone had any suggestions on what I might do to try and reduce these?
The output for my top model is below, with the largest SE being around 0.75.
Many thanks!

Output summary for CJS model
Name : Phi(~time + tsm + injury)p(~1)
Code: Select all
Npar :  8
-2lnL:  272.0516
AICc :  288.8428

Beta
                  estimate        se        lcl        ucl
Phi:(Intercept)  1.1913346 0.3825554  0.4415261  1.9411431
Phi:time2015     0.6786781 0.7595055 -0.8099528  2.1673090
Phi:time2016    -1.0533295 0.6412413 -2.3101624  0.2035035
Phi:time2017     0.4496663 0.7566217 -1.0333123  1.9326448
Phi:time2018    -0.6089459 0.7529656 -2.0847584  0.8668666
Phi:tsm          0.7522087 0.5328636 -0.2922040  1.7966214
Phi:injurynone  -1.9734351 0.4467178 -2.8490020 -1.0978681
p:(Intercept)    1.9795454 0.3193550  1.3536096  2.6054811
j.harv3y
 
Posts: 45
Joined: Mon Oct 08, 2018 4:45 am

Re: Large (ish) SE's

Postby jhines » Thu Mar 12, 2020 1:54 pm

Those estimates are on the logit scale, so they aren't bounded by (0-1). The Phi201x represent the differences between survival in the intercept year (2014) and the other years, on the logit scale. If you check the 'real' parameter estimates, the std. errors are probably smaller. Since the confidence interval on each of the differences covers zero, I'd try a model without the year-effect terms Phi(~tsm+injury)

If you want to see how the std. errors of the yearly estimates compare to each other, you could run the model, Phi(-1+time+tsm+injury), which will give the same likelihood as your model, but the betas will represent the intercepts (difference between estimate and zero) for each year.
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA


Return to analysis help

Who is online

Users browsing this forum: No registered users and 16 guests

cron