I'm amazingly rusty on basic statistics, but wouldn't something like the
following work?
Take the mean and standard deviation for the whole set. Then calculate
z scores for each value (essentially translate the values into their
deviations from the mean. Then, all you have to do is look for all
values whose z scores are too big or too small (both sides of the bell
curve).
I don't remember exactly how to do the math and I certainly don't know
how to code it in calc, but with the if, lookup, and vlookup, etc. it's
probably not that difficult.
Joe
Rod Engelsman wrote:
Lars D. Noodén wrote:
How would I go about using Calc's statistical functions to identify
extreme values in a data set ? e.g. in the following set, '98' is
way out of line with the other points.
1 2 3 4 5 6 7 8 11 12 13 15 16 18 21 29 34 36 44 98
It's been longer than I'm willing to admit since I've done any
statistics and never had any formal instruction. Would standard
deviation be useful for that? If so, how do I find the deviation for
individual points rather than the avg for the whole set?
-Lars
Lars Noodén ([EMAIL PROTECTED])
Patents are wrong for software but right for inventions. Write:
http://wwwdb.europarl.eu.int/ep6/owa/p_meps2.repartition?ilg=EN
Depends on what statistics you're trying to glean. There is a function
called trimmean() that lops off the bottom and top 5% of values before
calculating a mean. Also, in general, sometimes the median (the middle
value when the values are ranked) can be a more useful indicator of
central tendency than a straight mean. Economists use that a lot to
keep somebody like Bill Gates or John Rockefeller from screwing up the
average when looking at incomes or wealth.
Hmmm... reading your post again, I'm not exactly sure I understand
what you are asking. Do you want to calculate statistics in such a way
that the oddball values don't skew the results or are you looking for
a way to find the oddball values themselves to pinpoint anomalies?
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]