1995 年 16 巻 1.2 号 p. 19-36
Overdispersion data of chromosome-aberration rates from atomic-bomb survivors were analyzed by the quasilikelihood/pseudolikelihood (QL/PL) estimating equation method (Breslow 1990). The variance function was composed of two extra-binomial variations: one is an intra-individual correlation, the other is a variation from dosimetry error. This dose-error-variance component was derived using the quasi-structural method (Pierce, Stram, Vaeth and Schafer 1992) that incorporates large-scale external dosimetry information. Using the results of fitting the QL/PL estimating equation, the Wald test revealed that the dose-estimation error is a significant source of overdispersion.