Resetting original order after sorting out NaN values in the matrix

Hey everybody!

I have some problems regarding portfolio optimization and the extraction of weights. My problem can be summarized as fighting with NaNs constantly:

I have a matrix full of returns according to the following:

           Asset1     Asset2     Asset3 ....            Asset1000
1960_01     0.01       NaN        NaN
1960_02     0.015      NaN        0.02
1960_03    -0.02       NaN        0.1
...
2014_12

Now, some assets have many NaNs so I have written a code to keep only those Assets where I have all returns during the time period:

testifnan = all(~isnan(returns)); %this command returns a 1 if there are only NaNs in the column

new_ret = returns(:,testifnan); %returns a matrix with returns of stocks that do not have any NaN

Now, what I receive is a new matrix that is not a 660x1000 matrix as before, but only a 660x200 (800 stocks have only NaNs)

I do all this on a rolling basis, so after running the optimization code, I receive the set of optimal weights. I want these weights (of one optimization) in a 1x1000 matrix with a weight of zero if the asset had only NaNs in that time period.

Now, I created a matrix of zeros: weights = zeros(1,1000) and I want to update the weights in the original order..

I am really struggling with this and hope that somebody can help me out! Also, let me know whether you need more info!

Thank you so much!

Clemens

[the cyclist]

Just an FYI, you could also have done the NaN test in a slightly easier way:

testifnan = any(isnan(returns))

Answer 1

[the cyclist]

Use the index you created on the left-hand side:

weights(not(testifnan),:) = <...>