[R] Summary/Bootstrap for Design library's lrm function

[David Winsemius]

On Sep 26, 2009, at 5:11 PM, kkruse at uci.edu wrote:

> Can anyone tell me what I might be doing incorrectly for an ordinal
> logistic regression for lrm?
> I cannot get R(2.9.1)to run either summary nor will it let me  
> bootstrp to
> validate.

> ### Y is a 5 value measure with a range from 1-5, the independent
> variables are the same.  N=75 but when we knock out the NAs it comes  
> down
> to 51####
>
>> lrm(formula = Y ~ permemp + rev + gconec + scorpstat, data = data,
> na.action = na.delete, var.penalty = "simple")
>
> ## It will give me coefficients and residuals, but nothing else  
> really.
> When I try to enter "summary" it gives me this error message##
>
>> summary(bigassmall)
> Error in summary.Design(bigassmall) : could not find function "Varcov"

Frank has answered this question a couple of times in the last month.  
He has moved his active effort away from Design over to the rms  
package. In the process the Varcov function got left out of Design. He  
posted a replacement. I thought he was going to put it back into a  
fixed version, so the first thing I would check is to see if your  
version is outdated. If updating Hmisc and Design does not work, (and  
it did work for me), then see Frank's posting:

https://stat.ethz.ch/pipermail/r-help/2009-September/211306.html

... which also worked for me before I updated.

>
> ##So I thought I'd try to find a back door in, manually  
> bootstrapping to
> verify then getting values that way and I get this error message##
>
>> validate(bigassmall, method=boot, B=50)
> Error in validate.lrm(bigassmall, method = boot, B = 50) :
>  fit did not use x=T,y=T

?lrm

That seems to be a fairly explanatory error message. Looking at your  
call to lrm, which I infer from you code snippets and error messages  
was assigned to bigassmall, it certainly does not appear that you have  
set x=T and y=T.

>
> ########################################################
> Any clue as to what I'm doing wrong?

> any help would be much appreciated.
> Karl
> PhD Student, Political Science
> University of California at Irvine


David Winsemius, MD
Heritage Laboratories
West Hartford, CT