Re: [S] dummy.coef

Greg Arnold (G.Arnold@massey.ac.nz)
Wed, 18 Feb 1998 13:37:35 +1200


>I am analysing a reasonably large set of wool auction data to extract the
>economic value of different biological attributes of the wool sale lots. I
>have a model that includes factors and continuous variables. After some
>preliminary testing the best model looks like
>
>y = mu + Fi + Bj + ..m +m^2 + C:d...eij..
>
>It is the coefficients of the d within C's that I am most interested in.
>However, when I call dummy.coef after successfully running aov, I get the
>following message
>"Error in "%*%.default"(dm[[ii]], cfs): Number of columns of x should be
>the same as number of rows of y
>Dumped"

Ian,

I puzzled over this message in a different context, but there could be a
connection. The details of my problem are a bit hazy now, but the essence
was that I fitted an aov model to a subset of the data which happened to
omit one factor level completely.

Then I made predictions from the model.

Predict determined the size of the coef vector from levels(factor), which
was one greater than the number of coefficients in the model.

It could be that dummy.coef, like predict, is calculating the number of
coefs from the level attributes, which with your rather curious C:d term in
the model (do you mean C + C:d, which is equivalent to C/d?) could well
result in a mismatch of vector sizes.

Greg Arnold
Statistics, Social Science Building
Institute of Information Science and Technology
Massey University
P B 11 222
Palmerston Nth
New Zealand

g.arnold@massey.ac.nz

Phone +64 6 350 4254
Fax +64 6 350 2261

-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news