MATLAB and Simulink resources for Arduino, LEGO, and Raspberry Pi

# How can I do multiple regression (i = 1:15) in a loop?

Asked by Simon on 27 May 2013

I got a matrix A (1x15) with each cell containing 524 x 4 of data, and a Y matrix containing 524 x1 data. I was going for :

For i = 1:15
mdl{i} = LinearModel.fit(A{1,i},Y)
end

but it didn't work

per isakson on 28 May 2013

Any error message?

Simon on 28 May 2013

It seems to work but i end up having a 1x15 matrix of mdl, but all cells are empty... it's weird

Simon on 28 May 2013

No error message :/

## Products

No products are associated with this question.

Answer by the cyclist on 28 May 2013

This worked for me:

% Some pretend data
A = cell(1,15);
for i=1:15
A{i} = rand(524,4);
end
Y = rand(524,1);
% Do the fits
for i = 1:15
mdl{i} = LinearModel.fit(A{1,i},Y);
end

Simon on 29 May 2013

I mostly need the r squared, p value t-test and fisher. Is there a way to gather these without having to use mdl?

the cyclist on 29 May 2013

If the code I posted is giving completely empty results, I suggest you file a bug report.

An alternative to using LinearModel.fit is the regress() command.

Simon on 30 May 2013

I might report it eventually but i found out that this particular function generate an object and apparently i'm not supposed to have many objects in a single matrix. I manage to gather the Rsquared using Varname.Rsquared.Adjusted so that only answers pop out in a (1,15) matrix. The only problem i have here is that i don't have the combination variable names anymore and that might cause problem... If i try with 50 variables i won't be able to understand the answer! Thank you!!