Rolling stepwise regression within each regression window

I want to run a backward stepwise regression within each regression window. Basically, I have a Matrix Y(137x84) and a Matrix X(137x612). For each of the 84 Y's I want to check which of the 612 factors are significant within the a sliding window of 24 months and after that associate to the logical Matrix obtained the corresponding Xi. I wrote the following code to do that

b=zeros(K,N);
se=zeros(K,N);
pval=zeros(K,N);
inmodel=zeros(N,K);
win=24
for i=1:N

   dep_Step=Y(:,i);
  indep_Step=X(:,i);

for t=win:T
    x_roll_step=indep_Step(t-(win-1):t,:);
    y_roll_step=dep_Step(t-(win-1):t);
    [b(:,t),se(:,t),pval(:,t), inmodel(t,:), stats(t)]=stepwisefit(y_roll_step,x_roll_step);
end

i=i+1
end

cumulative=cumsum(inmodel,2);

A_i=cell(1,N);
match=zeros(T,K);
for i=1:N
    for j=1:K
        if inmodel(i,j)==1
            match(:,j)=X(:,j);%match 1 with the correspondent factor
            match( :, ~any(match,1) ) = []; %delete zero column from each matrix on each cell
         end
    A_i{i}=match; %store within the cell 

      end
      match=zeros(T,K);

       i=i+1 
  end

but the output A_i is just a structure with Matrix filled with zeros. Do you have any suggestion on what is wrong with above code?

Many thanks for your help.

[Federico Frisaldi]

Is anybody willing to help me this issue?