qq.loglogistic

I know of no such function.  Some authors have described PP and QQ 
plots for a arbitrary distributions.  I've experimented with PP plots 
and QQ plots for uniform, Student's t, chi-square and F distributions, 
but have not gotten much out of them.  The uniform plots failed to show 
adequate detail close to 0 or 1.  With the Student's t, chi-square, and 
F plots, the points in the tail were too far apart for me to easily 
judge what was happening.

	  However, I regularly use normal probability plots.  If I wanted to 
test a logistic fit, I might esimate the parameters, e.g. using 
"fitdistr" in library(MASS), then transform the data to probabilities 
using plogis, then further tranform the probabilities to normal scores 
using qnorm.  The I would make a normal probability plot.

	  If anyone else has relevant experience, whether consistent or 
conflicting with mine, I'd be pleased to hear about it.

	  spencer graves

qq.loglogistic

Thread (2 messages)