search

VincentWtrs / enetLTS

0 stars 0 forks source link

Sampling scheme for outcomes y 0/1 #1

Open VincentWtrs opened 5 years ago

VincentWtrs commented 5 years ago

When printing a contingency table for the outcomes that are indexed, i.e. the y[indx] as in the CStep() function, it is seen that the distribution is always exactly the same and always exactly as the 0/1 distribution in the original sample. However, upon further inspection, the actual rows composing these are indeed different, so some sampling is happening.

It is unclear to me how to explain this phenomenon at this moment.

VincentWtrs commented 5 years ago

Update: I don't know if it's truly a problem, less important. Effectively other rows are sampled, so this issue has lower priority.