Some notes on sampling tests with two variables.
成果类型：
Article
署名作者：
Pearson, ES
刊物名称：
BIOMETRIKA
ISSN/ISSBN：
0006-3444
发表日期：
1929
页码：
337360
关键词：
摘要：
R. A. Fisher''s test of goodness of fit of regression curves involves the assumption that the y arrays are homoscedastic normal curves. However, experimental sampling from a population with Type III curves ([beta]1 = .2; [beta]2=3.3) for y arrays and [sigma]x ranging from 5.7 to 4.6 agrees well with Fisher''s test. For markedly platykurtic or leptokurtic arrays Fisher''s test is not applicable. Methods are given for integrating f([theta]), which is involved in Fisher''s test. In experimental sampling from 2 non-normal populations the distribution of r agrees well with the theoretical values for a normal population. Fisher''s z transformation seems to represent the distribution of r adequately.