[R] exclude1 in summary.formula from Hmisc
Frank E Harrell Jr
f.harrell at vanderbilt.edu
Fri Jul 20 18:28:01 CEST 2007
david dav wrote:
> Here is a peace of the data and code :
>
> sex <-c(2,2,1,1,1,1,1,1,1,2,1,1,1,1,1,2,2,1,1,2,2)
> AGN <-
> c("C","C","C","C","C","A","A","C","B","B","C","C","C","C","C","C","B","B","C","C","C")
>
> X <- c(2,2,2,2,1,2,2,2,2,2,1,1,1,1,1,1,1,1,1,2,2)
>
> varqual <- data.frame(sex,AGN,X)
>
> desqual <- summary.formula(varqual$X ~., data = subset( varqual, select
> = -X), method = "reverse", overall = T, test = T, long = T, exclude1 = F)
>
> desqual doesn't show the results for sex ==1 as it is redundant.
> I also tried long =T wich didn't change anything here.
Oh yes. exclude1 is not an argument to summary.formula but is an
argument to the print, plot, and latex methods. So do print(desqual,
exclude1=FALSE, long=TRUE). Thanks for the reproducible example.
Note that you say summary( ) and don't need to type out summary.formula
>
> Bonus question if I may :
> This function is a little bit complex for me and I couldn't figure out
> how to get either Yates' continuity correction or Fisher Test instead of
> Chisquare. Does it ask a lot of program coding ?
Neither of those is recommended so they are not automatically supported.
But users can add their own test functions - see the help file for the
catTest argument to summary.formula. Fisher's test is conservative.
The Yates' continuity correction tries to mimic the conservatism of
Fisher's test. I don't like to get larger P-values when I can avoid it.
And the recommendations about worrying about the chi-square
approximation when some cell sizes are small are a bit overdone. Better
might be to use the likelihood ratio tests, and many other tests are
available.
Frank
>
> Regards.
>
> David
--
Frank E Harrell Jr Professor and Chair School of Medicine
Department of Biostatistics Vanderbilt University
More information about the R-help
mailing list