I'm running S+2000r1 under W2k. I have a rectangular dataframe of 23 columns, mostly numeric, a few string; and 233 rows-- but thereby hangs the rub -- as it's actually 234 ... well, I can add the la
Has anyone already written code for comparison of rotated EFA solutions across multiple groups? I have old things lying around that do a congruence coefficient; Cattell's modified s-index; and Kaiser
I'm away from my bibles of S+ over the Easter but want to finish a plot. I'm plotting a dependent against a fixed predictor with three levels. The data come from two experiments. The first was a pilo
A simple question. I have written a simple bit of S+ to give me confidence intervals for proportions based on the normal approximation based on: Gardner, M. J., S. B. Gardner and P. D. Winter (1989).
Thanks to Frank Harrell, Timothy J. Wade, Terry Therneau, Brad J. Bickerstaff and David Parkhurst who all sent me extremely useful answers. I'm much the wiser about exact and approximated binomial co
A number of people have asked me to post the responses I got to this. Here they are, with the permissions of the authors. Many thanks again to everyone. and my original question was: From here on sho
Aargh -- of course. So near and yet ... Here as an attachment, in case anyone really needs it (some people have been generous in thanking me so I guess I'm not completely barking!) Chris Chris Evans
Next low level question. I often use some of the ICCs (the reference I found most helpful to these was McGraw, K. O.,S. P. Wong (1996). ?Forming inferences about some intraclass correlation coefficie
I'm a researcher who often has to do my own statistics for various reasons so apologies if this and two other questions are pretty low level. I've just found the apply function and the array objects
Third and final low level question. I have to generate a lot of Kendall's W coefficients and to look at internal consistency with alpha, with particular interest in the corrected item-total correlati
Didn't seem like Kendall's W was around so here's a crude start I've cooked up. Suggestions for improvements, exact p values for small N and k and correct p values for data with ties all gratefully r
Thanks to everyone for their answers to this one. where twblack is my imported data frame, does exactly what I wanted, dumping the first three columns and taking the remaining 19 which were simple nu
A colleague has a small n (n=25) study of six variables. The variables are of good reliability and there is some a priori sense in which you would expect that three would covary and others also covar
Hi! I should have said that it was after varimax. However, it is precisely the problem you mention, or, to be more precise (I think) the problem that this is what you often see if you have "overextra
Thanks! I am a psychotherapist, not a mathematician so I guess I should just accept what you say, but I thought the difference was that a PCA assumes no specific "factors" but merely reorganises the
I'm sure this has been written up somewhere. Does anyone have a formula for the confidence interval of the Cohen type effect size parameter for a two group comparison (i.e. diff. of means/common s.d.
This seems a very dumb question, but what is the right way to test for a difference with paired data where the dependent is a trichotomy or higher polytomous variable? I know my McNemar's and Wilcoxo
I'm sure I'm missing the obvious here, apologies if I am. I am putting objects into a list using a function: add.to.parms <- function(name,object,long.name=" ",notes=" ") typically: add.to.parms("gri
... rest deleted ... Me being unclear I'm sure but many thanks to you and Brian Ripley, deparse(substitute(x)) was exactly what I wanted and I probably should have been able to imagine that any e.g.