corr returns NaN

I'm calculating the correlation coefficient between two vectors, using corr. It returns NaN, and I'm curious to know why that is the case.

c and d are the vectors.

>> [c d]

ans =

       1     1
       2     1
       2     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       2     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1
       1     1

Answer 1

[Sean de Wolski]

read the:

doc corr

and you'll see it's not what you want. Are you looking for :

doc corrcoef %?

[D]

Yes, thank you. But I still get

corrcoef(c,d)

ans =

1 NaN

NaN NaN

This seems to point at something in d as being an issue.

[Sean de Wolski]

It's because there's no variance/ standard deviation in the second column and thus in the correlation coefficient calculation when you divide by std or var (however it's implemented) you're in turn dividing zero by zero which yield nan.

[D]

Thank you. I missed that.

Answer 2

[Bio_Ing_Sapienza]

Maybe you have to ceck for "Inf" components in your array. Sometimes I was searching from Nan even if I had already checked them out. But there are other kind of Matlab numeric representation for results came from division that yields to too large numbers to be represented with floating number. So check also for these ones!