/sci/ - Science & Math

When you apply Chauvenets criterion and reject some data should standard deviation be recalculated using the rejected data or should the old values be used for further use?
(I am aware that old values can be used but you get bigger "error"). Books are split on this issue.

>>9315302
Find the point on the normal distribution that 98% of the area is to the right.

From the wikipedia pic, you'll want slightly more than 2 standard deviations before the average.

>>7366561
IQ scores (along with many other measurements, such as height or weight) can be plotted as a normal distribution - the bell curve. That means a very large portion of people (68%) are bunched within one standard deviation of the mean when the standard deviation is 15 points. If you have a clean bell curve of scores with the worst on the bottom and the best on the top, and you drew a line straight down the middle at 100, you'd find that 68% of your scores were from 85 to 115. The SD is a way of telling you just how unusual your score is relative to most people. A lower SD means more people are bunched around the mean, because on a normal distribution, 68% of people are always within one SD of the mean on either side. A higher score on a test with a higher SD isn't as impressive as one might initially think.