Why I can't run the corrcoef for matrix

Hi All, I am trying to run corrcoef to obtain the matrix of correlation coefficients and p-values but keep getting an error. I am using:

R = corrcoef (tbl);

(tbl is a 1050x5 table)

And this is the error I get:

Undefined function 'sum' for input arguments of type 'table'.

Error in cov (line 154) xc = bsxfun(@minus,x,sum(x,1)/m); % Remove mean

Error in corrcoef>correl (line 206) r = cov(x);

Error in corrcoef (line 89) r = correl(x);

Could someone please advise me what am I doing wrong and how to run this analysis?

Thanks in advance, Karina

[Karina Gutierrez]

Never mind, I just give it a second thought and realized that I had the input with the format of a table and not a matrix and this was creating the issue.

I created a matrix with the same data as the table and now it run just fine.

I guess I still don't understand the difference between a matrix and a table?

Cheers, Karina

Answer 1

[Steven Lord]

You can't compute the correlation coefficient of a table since a table could contain non-numeric data. Instead, you can compute the correlation coefficient of one or more variables in the table.

x = (1:10).';
y = x.^2;
t = table(x, y);
ycc = corrcoef(t.x, t.y)

If your variables can be concatenated into a matrix, you can pass the Variables from the table into corrcoef.

cc = corrcoef(t.Variables)

[Karina Gutierrez]

Thanks for your super fast response Steve, I concatenated the variables into a matrix and was able to run the correlation coefficient.

So just for clarification, is the main difference between a table and a matrix that tables can store non numerical values and matrices are only for numerical values? And if so, when doing any calculations is better to have matrices right?

Sorry, I am pretty new to Matlab and I still get easily confused.

Cheers,

Karina