Multivariate multiple linear regression?

[Aino]

Hi,

how can i do multivariate multiple linear regression with Matlab? I have four dependent (four Ys) and four independent (four Xs) variables. I believe that regress can do multiple regression (several Xs, one Y), and mvregress can do multivariate regression (several Ys, one X), but these doesn't help me much.

Thanks,
Aino

[dpb]

My reading of

doc mvgress

leads me to believe your second categorization (mult-Y, 1-X) is in
error. Read it again more carefully. I've not used it, but surely
looks to me like it'll handle it if you set the X/Y up correctly.

Failing that, you can always write the design matrix explicitly and use
backslash to solve the OLS equations assuming your design is
non-singular and of sufficient rank.

--

[Aino]

dpb <no...@non.net> wrote in message <kt5tvc$f6k$1...@speranza.aioe.org>...

Thank you for your quick reply! I read the instruction more carefully and you seem to be right, mvregress can do multivariate multiple regression. However, I still need to figure out how to construct the "design matrix".

-Aino

[dpb]

On 7/29/2013 10:10 AM, Aino wrote:
...

> Thank you for your quick reply! I read the instruction more carefully
> and you seem to be right, mvregress can do multivariate multiple
> regression. However, I still need to figure out how to construct the
> "design matrix".

Well, only you know what you want as the model...

--

[Aino]

dpb <no...@non.net> wrote in message <kt6iil$dum$1...@speranza.aioe.org>...

I think I thought that the design matrix was something mystical because of the eye-matrix that is part of the design matrix in the mvregress example. I found a function that creates a design matrix from a predictor matrix, x2fx. If 'interaction' is the right model for my case, then does this look right:

x=...;%prediction matrix, n x p
Y=...;%responces matrix, n x d
x=x2fx(x,'interaction');
[Yrow,Ycol]=size(Y);
X=cell(Yrow,1);
for j=1:Yrow
X{j}=[repmat(x(j,:),Ycol,1)];
end
[beta]=mvregress(X,Y);


-Aino