[R] Type III Sums of Squares

[Thomas Lumley]

On Mon, 28 Jan 2002, Moffet, Corey wrote:

> I am trying to reproduce an analysis done in SAS from which Type III Sums of
> squares are used. How can you get Type III sums of squares in an aov similar
> to what SAS provides?
>

You can get any set of sums of squares by fitting two aov models and
comparing them (it's a FAQ).  While I don't speak SAS fluently I believe
that the Type III sums of squares for a term A compare a full model to a
model without A but with everything else
eg
    full.model<-aov(Y~A+B+C,data=df)
    anova(full.model, update(full.model,.~.-A))


You can use
    drop1(full.model)
to get all of these at once.  By default this will not give Type III SS
for main effects in the presence of interactions but as these are usually
silly that's not a big problem.

If you really do want to test for a main effect in the presence of
interactions (which occasionally makes sense) you can specify a list of
specific terms to consider, eg:
    full.model2<-aov(Y~A+B*C,data=df)
    drop1(full.model2, scope=c("A","B"))



	-thomas

Thomas Lumley			Asst. Professor, Biostatistics
tlumley at u.washington.edu	University of Washington, Seattle

-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
r-help mailing list -- Read http://www.ci.tuwien.ac.at/~hornik/R/R-FAQ.html
Send "info", "help", or "[un]subscribe"
(in the "body", not the subject !)  To: r-help-request at stat.math.ethz.ch
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._