1

">cmatrix = covRob(X, corr=TRUE)"

Error in CovOgk(data, control = CovControlOgk(smrob = "s_mad", svrob = "qc")) : column with zero scale encountered in C function covOPW

I received the Error message above when trying to run X, 55x534 data matrix. It worked on another similar set of data but not this particular one that I am working on. It's quite strange as it's my first time seeing this error. Anyone has any idea what the error message is directed to?

1 Answer 1

0

covOPW does not have enough data to calculate the scale parameter, that is why scale is zero (for robust analysis, scale is the equivalent to variance in normal data, MAD would be an example of a robust scale parameter). You can have a look at the source code of covOPW.cpp

If you read up on the CovOgk algorithm (or other robust algorithms) you will find that often they take random or stratified samples of size 50%+ of the original data, to exclude outliers. In some case there may just not be enough data or the data is zero/NA.

In some robust estimators you can specify your expectation of the portion that is free of outliers. For example with CovMcd or CovMve you could set alpha = 0.9 instead of the default value alpha = 0.5. Than the samples will be larger in size and you may avoid this problem.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .