Problem with Portfolio Optimization, Mean Returns are not processed as vectors

Hey everybody,

I am trying to implement a classical portfolio optimization à la Markowitz with financial tooldbox. The data input is according to the following: (a 660x500 matrix)

          Asset1    Asset2    Asset3   Asset4 ......         Asset500
1960_01
1960_02
...
...
2014_12

the entries are monthly returns in that specific slot.

Now I try to perform the calculation of the efficient frontier using the Portfolio commands: First I cut the size of the set into intervals (later it will be in a for loop, for now I am trying to create a interval of the first 60 months):

ret_temp = ret(1:60,:); %subset of the dataset

.

Now, I am calculating the mean return and the var/cov matrix for each asset in this subpart:

mean_temp = nanmean(ret_temp); %returns a vector of mean returns of each asset cov_temp = nancov(ret_temp); %var/cov matrix of the sample period

now I can perform a portfolio optimization based on these inputs:

p = Portfolio; p = setAssetMoments(p,mean_temp,cov_temp); p = setDefaultConstraints(p); pwgt = estimateFrontier(p, 10);

Now, when running the code, the following Error appears:

Error using codename (line 18) Cannot set moments of asset returns.

Caused by: Error using Portfolio/parsearguments (line 101) Invalid AssetMean must be a scalar or a vector.

However, my asset means are clearly a vector and I cannot understand why this error is popping up..

Thank you very much in advance!

Clemens

Answer 1

[Brendan Hamm]

The problem is likely due to the presence of all NaNs in one of your return series, which results in a NaN even using nanmean.

rng('default')
X = rand(5,3); % Create a random matrix
X(2,:) = NaN; % Set an entire row to NaN
X(:,3) = NaN; % Set an entire column to NaN
nanmean(X)
ans =
  0.6219    0.6417       NaN

Notice nanmean will ignore the NaNs in a column, but if the entire column is a NaN you will have trouble. For this reason you will need to perform a little more data cleaning before hand.