>>I understand the MIN & MAX functions, but what is in the above sample of data, you added 6,8. I need to know what is a statistically correct way to reject these low numbers. How far off of the standard deviation is ok? Basically I want to get a reasonable average. These numbers skew things way off.
>
>you can use standard deviation if you have normal distribution (I am not sure of the term in english) of the data (ie the data basically fits the gauss bell)
> then you can calculate the mean and standard deviation
>and you can set the limits for your query as
>x between mean-n*deviation, mean+n*deviation)
>
>n=1 about 33 percent of the data
>n=2 about 67 percent
>n=3 97 percent
>n=3.5 99.something
>if my memory serves me right :)
>
Hope you don't mind if I amend these a little,
For two-sided interval, if we can assume a normal, n=1 is 68%, n=2 is 95%, n=3 is 99% (these are approx). You could probably use 2 std devs as a 'toss out any othe data' interval, but it depends on your situation.