Greetings ~
I need some assistance determining an appropriate approach to analyzing
multivariate binary response data, and how to do it in R.
The setting: Data from an assay, wherein 6-hours-post-fertilization zebrafish
embryos (n=20, say) are exposed in a vial to a chemical (replicated 3 times,
say), and 5 days later observed for the presence/absence (1/0) of defects in
several organ systems (yolk sac, jaw, snout, body axis, pericardial edema,
etc.) for each fish. The assay can be used as a screen (in which case a single
response 'any' is first analyzed) as well as for comparing the responses of
different classes of chemicals (a MANOVA-type analysis). Interest is focused on
where response(s) are occurring, any associations among responses, and
ultimately on possible biological mechanisms (the fish are tested for
behavioral activity prior to this assay, and then ground up to provide RNA for
microarray assays. A-lotta-data!).
What I *wish* I could do is something like glm(response.matrix ~ treat/vial,
family=binomial(logit), data=zf.dat) but I know this can't be done. I came
across the baymvb (Bayesian analysis of multivariate binary data) package in
the R contributed packages archives, but it is no longer supported and the
author is incommunicado. In the baymvb function the model is specified as
single.stacked.response ~ structure.factor + linear.model.terms. Huh? This
looks suspiciously similar to analyzing repeated measures data in SAS as a
univariate response with a split-plot structure (which forces the response
covariance matrix to have a compound symmetric structure). If this is what's
happening with this function it is definitely not appropriate. How about a GEE
approach (I'm not familiar with the literature, or with any of the R packages
that implement it)? Any other recommendations? (NB: just discovered the bayesm
package. Don't know if this will work for me or not.)
Any help would be greatly appreciated.
Derek Janszen, PhD
Sr Research Biostatistician
Computational Biology & Bioinformatics
Pacific Northwest National Laboratory
Richland, WA 99352 USA
derek.jans...@pnl.gov
[[alternative HTML version deleted]]
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
