Anonymous user
Talk:Statistics/Basic: Difference between revisions
→stddev slightly wrong
Line 26:
::To me, this conveys that we are to supposed to use the population stddev on a set of sampled data. And for that matter, the use of random numbers implies that we are talking about samples. --[[User:Rdm|Rdm]] 21:49, 2 July 2011 (UTC)
:: This is quite complicated. To put it simply, <math>1/(N-1)</math> is used when you need to infer the mean and stddev of an unknown distribution from a very small set of samples from it, while <math>1/N</math> is used when you take the available data set as fully representing the distribution or know exactly the mean of the distribution function. I chose to use the <math>1/N</math> formula is because 1) this task is merely for observing numbers, not doing any variance estimate; 2) it's simpler to derive and understand; 3) it does have its clearly defined meaning regarding how much the data vary, and the distribution is known to be uniform (though strictly speaking then the <math>\bar x</math> should not use the average of samples in this case, it should be exactly 0.5). In any event, when sample size goes up into hundreds, the difference between those two formulas become moot. --[[User:Ledrug|Ledrug]] 23:06, 2 July 2011 (UTC)
== a bit of duplication ==
| |||