Stata/SE or Stata/MP?

[Richard Williams]

I now face my periodic decision of whether to stick with SE or splurge and get 2 core MP. Quite frankly, IC would meet most of my needs. Does MP really make a huge speed difference? I will be doing some things with huge (for me) data sets, maybe a few million cases, but mostly for things like logistic and ordinal regression. And, I am starting to run more sem models. But, I've never been one of those people who say they have 2 billion cases and their stuff has been running for six months...

If I only get SE and realized I have made a terrible mistake, is it easy enough to upgrade to MP without taking a price hit, e.g. can you pay the difference between SE and MP or is the hit bigger than that?

[Bill Gould (StataCorp)]

To answer the last question first, you will save $25 if you upgrade directly from SE 13 to MP 14 rather than doing it in two steps.

In terms of most estimation models, MP/2 runs twice as fast as SE for the types of problems Richard describes. This is true for logistic and ordinal regression and most estimation commands, but not for SEM. SEM runs 20-30% faster, and that varies with model.

There's a white paper about MP performance; see http://www.stata.com/statamp/statamp.pdf