Compute correlation of two NaN-containing vectors

Hi all! If I need to find the correlation between two NaN-containing vectors by using corrcoef() or corr(), is it possible? How should I set up the parameters? and if not how is it possible to find the correlation between them? Thank you!

Here are two vectors:

Answer 1

[the cyclist]

The documentation for corrcoef has a section explicitly discussing different methods to handle NaN values. Click here to jump directly to that section.

Since you only have two vectors,

R = corrcoef(A,'rows','pairwise')

and

R = corrcoef(A,'rows','complete')

should give you identical results. (They could be different if you had more than 2 columns, depending on the pattern of NaN values.)

[Zhonghao Liao]

Thanks. I have one more question: what is the meaning of 'rows'? I mean I want to compute the correlation between these two column vectors. But when I try to input R = corrcoef(A,'columns','complete') it does not work and shows valid parameters are 'alpha' and 'rows'. Thanks!

[the cyclist]

The inputs are column vectors. But you are specifying that the rows of each column need to be complete. For example,

x = rand(5,1);
y = rand(5,1);
x(1) = nan;
y(2) = nan;
R = corrcoef(x,y,'rows','complete')

[Zhonghao Liao]

I see. Thank you!