Remove empty cells and values for correlations

Hi,

I have a 32x2 cell and I would like to do a correlation for the values in the first column with the values in the second column. However, there are blank cells in the first column. Could anyone help me find a way to do a correlation that excludes all the blank cells and their corresponding values in the second column?

Thanks!

[the cyclist]

"Blank cells" is not quite specific enough. Could you post a very small example (maybe 4x2 or so) that illustrates your input cell array? For example, do you mean

C = cell(4,2);

C{1,1} = 3;
C{1,2} = 4;
C{2,2} = 4;
C{3,1} = 6;
C{3,2} = 5;
C{4,1} = 4.1;
C{4,2} = 6.2;

where the 2nd row of the first column is an empty cell?

Or do you mean something else by "blank"?

[Thang Le]

I'm sorry for not being specific. An example would be:

[45] [20]

[16] [32]

[] [10]

[17] [6]

So (3,1) is an empty cell. So this is the same as your example, I think.

Answer 1

[Star Strider]

Does this do what you want?

CData = {45 20;  16 32;  [] 10;  17 6};
CE = cellfun(@isempty, CData(:,1));
Data = cell2mat(CData(~CE,:));
[R, P] = corrcoef(Data)

[Thang Le]

Thank you Star, Strider! This is indeed what I was looking for.

[Star Strider]

My pleasure!

[Thang Le]

Hi Star Strider, slight adjustment to my question: what if either column or both columns have [] (as opposed to just the first colmn as in my original question). How would I change the code to account for those possibilities? I've tried several things but usually got the Index exceeds matrix dimensions error.

[Star Strider]

It seems that none of the logical indexing that I find so useful with double arrays works with cell arrays (or at least it doesn’t work for me with cell arrays). Finding all the rows with a missing value was easy enough (my ‘CE’ variable). I finally resorted to a loop to collect only the data with complete rows:

CData = {45 20;  16 32;  [] 10;  17 6; 25 []};
CE = cellfun(@or, CData(:,1),  CData(:,2), 'Uni',0);
k2 = 0;
for k1 = 1:size(CData,1)
    if ~isempty(CE{k1})
        k2 = k2+1;
        CD(k2,:) = CData(k1,:);
    end
end
[R, P] = corrcoef(cell2mat(CD))

It works, but I’m sure there’s a more efficient way to do it.

[Thang Le]

Thank you again, Star Strider!

[Star Strider]

My pleasure!