A Frequency Curve Adapted to Variation in Percentage Occurrence

Publisher Website
Google Scholar
Abstract
It is shown that frequency distributions describing variation in percentage occurrence may often be fitted by the normal probability curve, following a transformation of scale from X to X 1( = prf -1 X) where X is the excess of the percentage occurrence of a given class over 50 %, and prf -1 X is the inverse probability integral of X defined by the relation X=prfX 1 [image] The logical basis is the assumption that back of the alternative categories there is a graded series of conditions, distributed according to the normal curve on a scale on which the same elementary factor has a uniform effect. The actual determination of the mean (a) and standard deviation (s) of the normal curve on the transformed scale can best be determined from the linear relation prf -1 p = 1/s(prf -l x[long dash]a) where p is the running sum of the frequencies between the median and the abscissa X. It is shown that the distributions of amount of white in the coat pattern in 10 inbred groups of guinea pigs fall under a simple common viewpoint on fitting by this method, although ranging between extreme positive and negative skewness.

This publication has 0 references indexed in Scilit: