[Ipmpack-users] help with IPMpack (Bruce Kendall)...confidence intervals

[Joachim uib]

With only 4 transitions I’d go for estimating the uncertainty of lambda based on the variability lying in the data.  As Bruce suggested by bootstrapping the underlying data (re-sampling individuals with repeats)…you get as many lambdas per transition as you specify re-sampling runs (which is arbitrary) and can hence calculate as many stochastic growth rates. Note, that the CI must then not be calculated via the standard error, beacause N is arbitrary. You’d need to go for quantiles. 
Another way to get the uncertainty from the underlying data into the IPM would be to run the vital rate regressions by bayesian inference. Instead of single model coefficients you’d get a distribution per coefficients and hence a distribution per matrix cell for all matrices and the entire IPM…and then of course a distribution of lambdas from which you can calculate a distribution of stochastic growth rates. With enough individuals I found both of these approaches to yield very comparable results.

Cheers,
Joachim




Sendt fra E-post for Windows 10

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/ipmpack-users/attachments/20160729/0b30f2db/attachment.html>