Dear All,
I’m stuck with something intractable. I have some raw time series, where the rows of NaNs at the start and end of the series vary. The stuff I need to do with this data only works for series not containing NaNs. The trouble is, the statistics I need are to be obtained for combinations of series. So if my data contains, say, 3 columns, then I want column 1 and 2, then column 1 and 3, and I have to store them (so I can do stuff with them later). Because the rows with NaN value at the end and beginning of the three column varies, the combination of column 1 and column 2 will be of different size from the combination of column 1 with 3. The way to do this, I believe, is to use a cell array, which caters for something like that. So if my code is:
[n,l] = size(data); %find length of data (raw, including NaN rows)
B = cell(nCols-1,1); %preallocate a cell to store matrices of different row length
for ii= 1:nCols
%loop to pair match-wise rows without adjacent NaNs (two columns at a time)
for i=1:n;
if sum(isnan(data(i,:)))>=1
i=i+1;
else
break
end
end
B{ii}=data(i:n,:);
end
The inner loop (from the second ‘for’ to the second last ‘end’)works (I tried it on two columns of data, it does the job-it truncates the sample (for both) at the first row where both vectors contain data.
The column index in the RHS of the second last line is obviously wrong-what I would like matlab to do is to pick column 1 and column 2, truncate them at the appropriate length (starting with the 1st adjacent pair of rows not containing NaNs, and storing them in the cell, then for ii = 2, I would like matlab to take column 1 and column 3, do the same operation, and stick the resulting pair of columns in the next cell, and so on (here, for expositional clarity there’s only 3 time series, in reality there’s a large number of them.
What to do? I’m at a loss.
Many thanks in advance.
