[R] Error in ANOVA for model comparison

[John Fox]

Dear Rosario,

Because of missing data in the additional variable PHt, the two models
weren't fit to the same subset of valid observations -- the default in lm()
is to use complete cases for the variables in the model. 

A mechanical solution is to use na.omit() to filter your data set, only for
the variables you intend to use, to produce a data set with no NAs. Then
you'll fit each model to a consistent subset of valid cases. 

Of course, if you have a substantial amount of missing data, complete-case
analysis is probably a poor strategy.

I hope this helps,
 John

--------------------------------
John Fox
Senator William McMaster
  Professor of Social Statistics
Department of Sociology
McMaster University
Hamilton, Ontario, Canada
http://socserv.mcmaster.ca/jfox




> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
> On Behalf Of Rosario Garcia Gil
> Sent: January-21-11 9:31 AM
> To: r-help at r-project.org
> Subject: [R] Error in ANOVA for model comparison
> 
> Hello
> 
> I am trying to compare two models using anova(), however I get a message
> error (see below).
> In the net I only found some information on certain library(car) for
> which one should use anova with A capital letter (Anova instead of
> anova), but I could not find car library as it says it does not exist.
> 
> 
> > Model <- lm(interceptG ~ SW + TSC + FSC + PF + SlopeG + K,
> data=AllTrait)
> > Model1 <- lm(interceptG ~ SW + TSC + FSC + PF + SlopeG + PHt,
> data=AllTrait)
> 
> Error in anova.lmlist(object, ...) :
>   models were not all fitted to the same size of dataset
> 
> I have NA in the datafile, should that be the problem?
> 
> Kind regards and thanks in advance
> Rosario
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-
> guide.html
> and provide commented, minimal, self-contained, reproducible code.