Incorrect dimensions for matrix multiplication

%Load the data
load data.mat;

%Extracting first 500 points
z = [y(1:500) u(1:500)];

%Plotting the data
figure(1)
idplot(z);

%Using sampling interval of 80ms
figure(1)
idplot(z, 1:500, 0.08);

%Reserving the remaining data
zr = [y(501:1000) u(501:1000)];

%Removing constant levels and making data zero mean
z = dtrend(z);
zr = dtrend(zr);

%Loss function calaculation
V = arxstruc(z, zr, struc(2,2, 1:10));
value = V(1,:);
nk_value = min(value);

V = arxstruc(z, zr, struc(1:5, 1:5, 3));

%Searching for settled-in point
nns = selstruc(V);
i=1;

%Model structure
for na = 1:5
    for nb = 1:5
        val(i,:) = na+nb;
        th = arx(z,[na, nb, 3]);
        mse(i,:) = [th.Report.Fit.MSE];
        a = th.a;
        b = th.b;
        ysim = idsim(zr(:, 2), th);
        i=i+1;
        matx = [val mse];        
             
        A = matx;                                                      
        [A1u,~,idx] = unique(A(:,1));                                   
        MinVals = accumarray(idx, A(:,2), [], @(x)min(x));             
        Result = [A1u, MinVals];                                       
        
        m_one = Result(:,1);
        m_two = Result(:,2);
        
        figure(2)
        plot(m_one, m_two)
        title('Complexity vs Loss Function');
        xlabel('Complexity');
        ylabel('Loss Function');
       

    end
end

%A-parameter and B-parameter
th = arx(z,[4 4 3]);
th = sett(th, 0.08);
present(th);
A=th.a
B=th.b


%Least Square Estimation
i = 1;
for k=6:length(z)
    phi(i,1:5) = [-z(k-1,1) -z(k-2,1) z(k-3,2) z(k-4,2) z(k-5,2)];
    i = i+1;
end

Y = z(6:end,1);
L = inv(phi'*phi)*phi*Y;

Getting an error that says 'Incorrect dimensions for matrix multiplication.'

Can anyone explain why this error comes up?

[Walter Roberson]

This code requires idplot(), which existing in the MATLAB 3 (1987) to MATLAB 5 (1997) timeframe and which was removed from MATLAB before R13. A copy of the source appears to be at http://research.jisao.washington.edu/vimont_matlab/System/idplot.html

Answer 1

[Adarsh Ghimire]

L = inv(phi ' * phi ) * phi' * Y;

your phi matrix dimension has to transposed to multiply with y. As i went through your code, I found that

dimensions:

phi = (z-6) x 6

and y = (z-6)x1

so transposing your phi and multiply with y will get you 6x1 matrix which can multiply with inverse result of 6x6 so you get 6x1 matrix as your L.

[Gautami Golani]

Sir,

I have changed that part as you suggested, but I am still getting the same error stating that the dimensions are incorrect for matrix multiplication.

Thanks and regards

[Adarsh Ghimire]

can you tell which at which line is that multiplication error ?

[Gautami Golani]

It is at the last line of the code where I need to compute 'L'.

[Adarsh Ghimire]

Use debugging at that line at determine the size of those phi and y, check if the dimensions are well matching for multiplication to be carried.

It may be due to degenrative nature of phi' x phi matrix, so try this one.

L = pinv(phi'*phi)*phi'*Y;

[Gautami Golani]

I found out the size of phi and Y. The sze of phi is 500 5 and that of Y is 495 1.i

[Adarsh Ghimire]

then you manage the dimension