askel50 on 05 November 2012
| ishiki said: If you're throwing out outliers you have to throw some 100's too. If you just throw out the bottom scores that's dumb. And you're just inflating the score. Or just not do this at all. |
That's not how it works. The principle is that values that differ too much from the mean are anomalous, you can keep track of them but you shouldn't use them. If 99 persons say 10 and 1 person says 0, the effective mean should be 10 because that one person is an anomaly, is not normal. You don't take out 0 or 10 because they are extreme, you take them out because they are too big or too small in relation to the expected value. If the mean is high you take out small values, if the mean is low you take out high values.
