Large-scale sample surveys cause the practical problem that any statistical testsfor the survey data must reject null hypotheses easily. If sociologists receive the results of the statistical tests with no consideration, they should make an error toregard trivial characters of the population as important features. This paper applies‘the generalized x2 goodness-of-fit test’ to the data of JGSS-2002, and show one of themethods to solve the problem. The generalized x2 goodness-of-fit test was developed in medical data analyses, but it should be useful in sociological data analyses. Because the method can test the null hypothesis that non-zero amount of lack of fit is present, and check whether there is ‘the large amount’ of divergence from a model or not. The result of the application of the method to JGSS-2002 data was very good, because the method could clarify whether the large amount of divergence was present. If we can establish the appropriate amount of divergence for null hypotheses, thegeneralized x2 goodness-of-fit test will be a very useful instrument for the analyses oflarge-scale samples.
展开▼
