R help - Jul 2005 - Design: predict.lrm does not recognise lrm.fit object

Hello

I'm using logistic regression from the Design library (lrm), then fastbw to
undertake a backward selection and create a reduced model, before trying to
make predictions against an independent set of data using predict.lrm with
the reduced model.  I wouldn't normally use this method, but I'm
contrasting the results with an AIC/MMI approach.  The script contains:

# Determine full logistic regression
lrm_logist = lrm(PresAbs ~ Size + X2ndpc + soil + AAR + tjan.jun,
data=training)
# Backward selection of variables in model
lrm_stp = fastbw(lrm_logist, rule="p", sls=0.05)
# Fit reduced model
lrm_reduced = lrm.fit(training[,lrm_stp$parms.kept[-1]], training$PresAbs)
# Predict using parameters from reduced model against a new dataset
predict(lrm_reduced, testing, type="fitted.ind")

It is the last command that fails, reporting the error:
Error in getOldDesign(fit): fit was not created with a Design library
fitting function.

On further investigation, the class of the object from lrm.fit is only
"lrm", whereas predict.lrm seems to be expecting objects with the
class
"lrm", "Design", "glm", judging from the examples
on the predict.lrm help
page.  Many of the attributes differ also.  Does anyone know a simple
work-around for this problem?

Many thanks
Roy

Roy Sanderson wrote:
> Hello
> 
> I'm using logistic regression from the Design library (lrm), then
fastbw to
> undertake a backward selection and create a reduced model, before trying to
> make predictions against an independent set of data using predict.lrm with
> the reduced model.  I wouldn't normally use this method, but I'm
> contrasting the results with an AIC/MMI approach.  The script contains:
> 
> # Determine full logistic regression
> lrm_logist = lrm(PresAbs ~ Size + X2ndpc + soil + AAR + tjan.jun,
> data=training)
> # Backward selection of variables in model
> lrm_stp = fastbw(lrm_logist, rule="p", sls=0.05)
> # Fit reduced model
> lrm_reduced = lrm.fit(training[,lrm_stp$parms.kept[-1]], training$PresAbs)
> # Predict using parameters from reduced model against a new dataset
> predict(lrm_reduced, testing, type="fitted.ind")
lrm.fit is used internally by lrm and related functions.  You'll need to 
use lrm.  Take lrm_stp$names.kept and create a formula - here is a first 
stab:

form <- as.formula(paste('PresAbs ~', 
paste(lrm_stp$names.kept,collapse='+')))
lrm_reduced <- lrm(form, ...)

This only works because you are forcing everything to be linear (highly 
  unlikely) and are not allowing any interactions.

Remember that is you use resampling to validate the model you need to 
start with the full list of candidate variables for each resample, 
having validate or calibrate functions repeat fastbw internally.

Frank
> 
> It is the last command that fails, reporting the error:
> Error in getOldDesign(fit): fit was not created with a Design library
> fitting function.
> 
> On further investigation, the class of the object from lrm.fit is only
> "lrm", whereas predict.lrm seems to be expecting objects with the
class
> "lrm", "Design", "glm", judging from the
examples on the predict.lrm help
> page.  Many of the attributes differ also.  Does anyone know a simple
> work-around for this problem?
> 
> Many thanks
> Roy
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
> 

-- 
Frank E Harrell Jr   Professor and Chair           School of Medicine
                      Department of Biostatistics   Vanderbilt University