I'm helping a colleague who is trying to generate confidence intervals
around a point estimate of haldanes (a measure of evolutionary change
resulting from dividing the difference in two means by the standard
deviation). The data is analyzed with a repeated measures generalized
linear model: multiple individuals from each population (the 'treatment')
that we are comparing, and for each individual there are multiple (hence RM
design) yes-no responses (hence generalized model.) We previously used a
bootstrap approach to generate confidence intervals with a similar dataset
that did not include repeated measures by extracting the effect sizes with
the 'effects' library. However, we are having trouble figuring out a way of
extracting the effect sizes (estimated marginal means) from a generalized
estimating equation in the 'geepack' library for the repeated measures
data. I have no experience with GEE or extracting marginal means, but I'm
comfortable working in R so I'm just looking for help from anyone who might
know what function I could use to get the necessary values to bootstrap.
Can anyone offer advice? Below is my colleague's description of the
problem, in case that clarifies it more. Thanks!



Brad Carlson



"We are attempting to bootstrap haldanes (a measure of trait divergence)
calculated from results of a repeated measures generalized lineal model
(RGZM).  To bootstrap haldanes we need to run the model multiple times
(e.g. 500) using a randomized subset of the raw data, then calculate a
corresponding haldane.  The output that we need includes the sum of squared
deviations (sum of squares) associated with each treatment (population) or
the variance, or something we can use to get to it (i.e. standard error or
standard deviation).  We can run the model, but we cannot figure out how to
get the necessary stats.  We have also not been able to duplicate the
results from SPSS repeated measures generalized linear model in GEE in R.
The means (from the original full data set) are very close, but not exact.

We were successful in bootstrapping haldanes calculated from a simpler
generalized linear model, which was not repeated and didn't require us to
use GEE.

Does someone understand the GEE sub routine well enough to help us out? "

-- 

Bradley Evan Carlson
PhD Candidate
Intercollege Graduate Degree Program in Ecology
The Pennsylvania State University
University Park, PA 16802

Email: [email protected]
http://homes.bio.psu.edu/people/faculty/langkilde/index_files/carlson.htm
https://sites.google.com/site/bradleyecarlson/home

        [[alternative HTML version deleted]]

_______________________________________________
R-sig-ecology mailing list
[email protected]
https://stat.ethz.ch/mailman/listinfo/r-sig-ecology

Reply via email to