Why this message error?

n=size(prices,2);
M_coint=zeros(n,n);
for i=1:n;
    j=1:n;
    M_coint(i,j)=cadf(prices(:,i),prices(:,j),0,1);
    M_coint(i,j)=M_coint.adf;
end

message error: The following error occurred converting from struct to double: Error using double Conversion to double from struct is not possible.

[fede]

cadf give the following results: ans =

    alpha: -0.0199
      adf: -3.0082
     crit: [6x1 double]
     nlag: 1
     nvar: 1
     meth: 'cadf'
I want to pull out the result of adf

Answer 1

[Stephen23]

cadf does not seem to be an inbuilt MATLAB function, and you give no information about how it is defined or the output that it returns. And you also don't give us the complete error message (which tell us the line), so we have to also guess where this error is occurring in your code.

I like guessing games, so lets try.

First you define

M_coint=zeros(n,n);

which means M_coint is a matrix of class double. Then if we assume that cadf returns a structure, then the line

M_coint(i,j)=cadf(...)

will try to allocate that structure to M_coint.

Ouch! Allocating a structure to a double is an error! It does not work.

[fede]

cadf give the following results: ans =

    alpha: -0.0199
      adf: -3.0082
     crit: [6x1 double]
     nlag: 1
     nvar: 1
     meth: 'cadf'
I want to pull out the result of adf

[Stephen23]

Perhaps you meant to do something like this:

    tmp = cadf(prices(:,i),prices(:,j),0,1);
    M_coint(i,j) = tmp.adf;

Or alternatively you could save the complete output in a non-scalar structure, which means that you will have all of the data that cadf returns:

N = size(prices,2);
for n = N:-1:1;
    S_coint(n) = cadf(prices(:,n),prices,0,1);
end

This also makes accessing the data easy, try:

[S_coint.adf]
{S_coint.method}
horzcat(S_coint.crit)

[fede]

mmmm ok the error message disappear, but I have not that I want. I have a matrix, in which in the each columns I have the single stocks, and in the row the time serie price. For example

IBM JPM C p11 p12 p13 p21 p22 p23 .................... pn1 pn2 pn3

I want obtain a output as a var-cov matrix, but in place of the covariance the score of cointegration test, as the following table.

      ibm    jpm  c
ibm   -     -3.2 -4.2
jpm -1.2     -   -2.1
c      -2.3 -4.3     -

I can not generate a code suitable to my needs

[Stephen23]

This is a different topic to your original question. Please search for solutions to this new topic in the documentation, on the internet, or in this forum.