2006 Volume 21 Issue 1 Pages 33-48
Quantitative sociological researches in Japan have not tried to apply the simultaneous analysis employing interaction variables but have used the comparison of separated analyses among sub-samples by a model with the same independent variables, when the research task accompanies multiple-comparison. However, the comparison of separated analyses tends to be influenced by subjective eyeballing judgment about differences among sub-samples. How to carry out multiple-comparison by the simultaneous analysis using logit models on educational achievement was examined and illustrated. As the simultaneous analysis model concerning educational attainment, ordered logit model which is more parsimonious than transitional model and multinominal logit model in terms of the number of parameters is advantageous to multiple-comparison. Efforts to explore the possibility of simultaneous analysis should be required for not only logit models but other kinds of models.