How can I create an array of LinearModel objects

I am modeling a system with multiple inputs (X) and multiple outputs (Y). I´m using stepwiselm to model each output Y(i) separately:

mdl = stepwiselm (X,Y(:,i),'Upper','linear','PEnter',0.03,'PRemove',0.05)

it works fine, but when I try to create a loop and store all models:

mdl(i) = stepwiselm (X,Y(:,i),'Upper','linear','PEnter',0.03,'PRemove',0.05)

there is an error:

Error using classreg.regr.FitObject/subsasgn (line 738) Assignment using () is not allowed for a FitObject.

I have tried to concatenate models in several ways:

prove=[mdl1;mdl2]

Error using classreg.learning.internal.DisallowVectorOps/throwNoCatError (line57) Concatenation of LinearModel objects is not allowed. Use a cell array to contain multiple objects.

Error in classreg.learning.internal.DisallowVectorOps/vertcat (line 49) function a = vertcat(this,varargin), throwNoCatError(this); end

prove=cell(mdl1,mdl2)

Error using cell Conversion to double from LinearModel is not possible.

prove=cell{mdl1,mdl2}

The result of calling the static method or constructor "cell" cannot be indexed with cell array indexing.

but with no result.

I can store one model in a different variable:

mdl1=mdl

but I need to store a different number of models to be used later, and it's mandatory that they can be indexed.

I have tried to overcome this by storing some parts of the LinearModel objects in separate arrays

coeficients(i,1)=mdl.Coefficients.Estimate(1);
for j=2:1:mdl.NumCoefficients
    for k=1:1:M
        if strcmp(cell2mat(mdl.CoefficientNames(j)), cell2mat(NomVarDat(k)))
           coeficients (i,k+1)= mdl.Coefficients.Estimate(j); 
        end
    end    
end
fitted(i,:)=mdl.Fitted;
residus(i,:)=mdl.Residuals.Raw;

but then I want to use some LinearModel methods to evaluate new variables, ....

So, the question is: How can I create a indexable group of models (LinearModel objects)?

Thank You

Answer 1

[Steven Lord]

Store them in a cell array.

[Manuel Lamich]

Thank you for your answer, I have tested

prove=cell(mdl)

and Matlab returned: Error using cell Conversion to cell from LinearModel is not possible.

prove=cell{mdl}

The result of calling the static method or constructor "cell" cannot be indexed with cell array indexing.

finally it works in a simple way:

prove ={mdl;mdl2}

[Steven Lord]

That's one way; another is to preallocate the cell array and then use indexed assignment to store the LinearModels in each cell.

 C = cell(1, 10);
 for k = 1:10
   C{k} = magic(k+2);
 end
 X = C{5} % X will be magic(7)