Re: Type III sums of Squares.

Bill Venables (wvenable@attunga.stats.adelaide.edu.au)
02 Aug 1994 15:19:51 GMT


>>>>> "Ben" == Ben Lyons <lyons@stat.orst.edu> writes:

Ben> Help!!

Ben> Does anybody know an easy way of getting SAS like type III sums of
Ben> squares in S+?

This is almost an FAQ. The two rules are

1. Use contr.poly, contr.helmert or contr.sum contrast options for factors,
but *not* contr.treatment,

2. Use drop1(.,.) with a specific scope argument:

eg
> options(contrasts=c("contr.helmert", "contr.poly") # the default, but
> # make sure
> fm <- aov(y ~ A*B*C + x, data=my.data) # or whatever...
> drop1(fm, . ~ .)

Disclaimer: My considered opinion is that Type III sums of squares for
non-marginal terms, such as main effects with interactions involving them
are still present in the model, are very often of no interest whatever.

Bill

--
___________________________________________________________________________
Bill Venables, Department of Statistics,          Telephone: +61 8 303 3026 
The University of Adelaide,                       Facsimile: +61 8 232 5670
South AUSTRALIA.     5005.            Email: venables@stats.adelaide.edu.au