Re: [R] Multiple comparison for circular data

From: Spencer Graves <>
Date: Sat 18 Feb 2006 - 14:07:00 EST

          I just got 141 hits from 'RSiteSearch("circular data")'. From this, I quickly learned of two contributed packages for circular data: "circular" and "CircStats". Both seem to be current, and both are maintained by the same person.

          If you can compute a likelihood for a particular model, then you can test for equality of parameters between two data sets as follows:

	  logLik1 = log(likelihood for data set 1 fit by itself.
	  logLik2 = log(likelihood for data set 2 fit by itself.
	  logLik12= log(likelihood for data sets 1 and 2 fit as one.

	  log(likelihood ratio) = (logLik1+logLik2-logLik12).

	  If k parameters are fit for each model, then 2*log(likelihood ratio) 
should be approximately chi-square(k). From this you can get a p-value to compare any pair of data sets. Then use Bonferroni, multiplying each p-value by, say, (k-1). Any p-value(s) below your Bonferroni-adjusted threshold could be claimed to be significantly different.

          If this is not adequate, I suggest you download both packages (if you have not already), try them, and then send a question to the maintainer.   If you would like more help from the listserve, I suggest you PLEASE do read the posting guide! "". Also, please install the "fortunes" package and read especially fortune('posting').

	  hope this helps.
	  spencer graves

ecoinfo wrote:

> Hi All,
> Does anyone know how to compare the means of several circular-data
> samplings? Any related websites, references and softwares?
> Thanks
> Xiaohua
> --
> Xiaohua Dai, Dr.
> [[alternative HTML version deleted]]
> ______________________________________________
> mailing list
> PLEASE do read the posting guide! mailing list PLEASE do read the posting guide! Received on Sat Feb 18 14:13:08 2006

This archive was generated by hypermail 2.1.8 : Mon 20 Feb 2006 - 14:08:41 EST