"RenE J.V. Bertin" <rjvbertin at hotmail.com> writes:
> Let's say I have a within-subject experiment with 2 observables, obs1
and ob2 and 2 independent factors, fac1 and fac2.
>
> I can do
>
> summary( aov( obs1~fac1*fac2 + Error(Subject/(fac1*fac2)) ) )
> summary( aov( obs2~fac1*fac2 + Error(Subject/(fac1*fac2)) ) )
>
> to test the 2 observables separately.
>
> > summary( fit<-manova( cbind(obs1,obs2)~fac1*fac2 +
Error(Subject/(fac1*fac2)) ) )
>
> gives results that seem sensible, but that I suppose represent the
> effects on obs1 and obs2 combined. The example for summary.manova
> shows that > summary.aov(fit) would print the effects separately for
> each observable. However, I get the message
>
> Error in 1:object$rank : NA/NaN argument
>
> and
> > class(fit)
> [1] "aovlist" "listof"
>
>
> I suppose that this means that I shouldn't want to do this sort of
> test, and rather look at each observable separately. The question
> is: why? Is there a simple explanation comprehensible for the
> non-statistician?
This is not my specialty, but could it be that you have an empty error
stratum? I.e. if Subject:fac1:fac2 is the entire data set, then there
is no DF for the residual error. (If so, try Error(Subject/(fac1+fac2)) )
--
O__ ---- Peter Dalgaard Blegdamsvej 3
c/ /'_ --- Dept. of Biostatistics 2200 Cph. N
(*) \(*) -- University of Copenhagen Denmark Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk) FAX: (+45) 35327907
-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-.-
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
_._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._._