1997 年 62 巻 502 号 p. 65-70
The purpose of this paper is to investigate particular problems from the application of the logistic analysis to the case of multiple responses for the thermal sensation votes, and to find general solutions of them. The authors suggest that the logistic function should apply to the probability of a vote corresponding to environmental temperature, because the logistic model is considerably simpler than other binary models. Observed the probability of either zero or 100 per cent, the modification rule of Berkson is found to be efficient. The test of Cochran-Armitage is recommended using on extremely small or distorted samples before the regression analysis. The analyzer should choose between two types of the graph of logistic curves according to aims for the thermal sensation studies.