michalrichards9

3 ms·

1. 1.▲
   by michalrichards9 5y ago

   The statistics in this paper are... not robust IMHO. They do so many comparisons between the groups and never mention any multiple comparisons correction. Interesting idea, I am not convinced right now - but let's see if it reproduces.
2. 2.▲
   by michalrichards9 7y ago

   welp, that's the kind of code an idiot would have on his luggage
3. 3.▲
   by michalrichards9 7y ago

   My little rule of thumb is to inherit if the new implementation is a 'pure' over the inherited class. That is, I'm not introducing any _new_ changes to the state of the class.
4. 4.▲
   by michalrichards9 7y ago

   Thanks. I think you are furthering the point of prediction vs. modeling, right? You can, after all, get a confidence rating for a model
5. 5.▲
   by michalrichards9 7y ago

   I am not an expert and am still reading thru the article, but why is it such a strong dichotomy? Don't all predictive algorithm also assume a data model? for example aren't hidden Markov models, by assuming constant transition pr