[SciPy-user] scipy.stats.samplevar, samplestd and stderr and sem - redundant?

josef.pktd@gmai... josef.pktd@gmai...
Thu Dec 18 12:40:39 CST 2008


Following up on the discussion on statistical functions, I checked
again scipy stats:

There are two more functions: samplevar, samplestd, which are just
variance and standard deviation with hardcoded denominator n
(confusing terminology to call them sample...). I think they should
also be deprecated and removed.

stderr and sem have not very clear description and they both give the
same results, essentially

sqrt( sum( (deviations)**2 )/(n-1.0)/n )

>From the description I would infer that they use different
denominators, but that is not the case.

I'm not sure what the correct denominators are for this case, but one
of the two function, could be deleted, I like `sem` - "standard error
of mean" better than the generic `stderr`.

Josef


More information about the SciPy-user mailing list