Alex- data should only be discarded in curve fitting if it is so improbable as to not reasonably be expected. Many if not most statisticians/scientists would say NO data should EVER be thrown away. I agree with that school. But either way, 10% would be ludicrous by anyone's standards.
One standard deviation = 68%
Two standard deviations = 95%
1.6 standard deviations would contain roughly 90% of the data. You're saying to throw away everything outside of 1.6 standard deviations of data? The only reason data is discarded anyway is in curve-fitting and when you have more than 100 million points you don't have to worry that you can't fit the data properly. Then you have a gaussian curve with no ends and you'd be throwing away 10 million + points of data. Throw away millions of points and THEN you would have misleading data.
Last edited by Colin at Nov 15 2002, 04:34 AM
|