Used CROSSVALIND to Randomize rows with numerical only but not NAN

Dear all, I have a column double vector which consist 0,1 and NAN. The big idea was to assign the column vector into either one of the group, group 1, group 2,... group 10. The group assignment was realized using CROSSVALIND. However, I only want to assign iff the element is 0 or 1.

For example.

Assume the column double vector

xx = [NaN;NaN;NaN;1;1;;0;0;1;NaN]

Thus, the expected output using CRASSVALIND will be something will

group = [NAN;NAN;NAN;3;4;1;7;4;NAN]

Simply plug in the xx vector as following

Group = crossvalind('Kfold',xx,10);

produce the following error

Error using accumarray
First input SUBS must contain positive integer subscripts.

Thus, the following dirty work is propose

load('xx');
yyy =find (~isnan(xx));
Group = crossvalind('Kfold',yyy ,10);
newGroup =nan (length(xx),1);
for i=1:length(yyy)
newGroup(yyy(i))=  Group(i);  
end

However, I wanted to know if MATLAB allow better ways to achieve the same goal?

I attached together the MAT file containing the xx vector together with this thread Thanks in advance for the time entertaining this thread.

Answer 1

[Walter Roberson]

Not a "better" way, but correcting your code and optimizing slightly:

load('xx');
mask = ~isnan(xx);
Group = crossvalind('Kfold', xx(mask), 10);
newGroup = nan(length(xx),1);
newGroup(mask) =  Group;

[balandong]

Hi Walter, Thanks for input, I really appreciate it