I have come across this calculation in a script involved in some principal component analysis:
ddata = data'*data; ddata = max(ddata,ddata');
The resulting matrix ddata is used where I would normally expect a covariance matrix. data is a matrix containing data vectors as its rows. This is not how I would normally calculate a covariance estimate, so I was wondering - is this some kind of robust covariance estimate? I have tried to look it up, but I only found a so-called sandwich estimate, which does not seem to be this.