A new method of experimental sampling illustrated on certain non normal populations
成果类型：
Article
署名作者：
Hey, GB
刊物名称：
BIOMETRIKA
ISSN/ISSBN：
0006-3444
发表日期：
1938
页码：
6880
关键词：
摘要：
Sampling is effected by punching both original data and random numbers on to Hollerith cards. Sorting and tabulating machines take random samples and compute the various statistical parameters for each sample with great speed. The method is used to demonstrate on certain non-normal populations, both actual and artificial, that the distribution of the estimates of the correlation and regression coefficients in samples from a bi-variate population where the correlation is high, is not seriously affected by departures of the original population from normality. This also applies to Fisher''s z in conditions corresponding accurately with those in a Randomised Block expt. The method can easily be extended to the distribution of other parameters.