The AID computer programme, used to predict adoption of family planning in Koyang

Publisher Website
Google Scholar
Abstract
A highly useful computer programme has been developed for predicting values of a dependent variable. Basically a sequential analysis of variance, it creates a tree of two-way splits of the sample. Each split maximizes the reduction of unexplained variance in the dependent variable. The programme is remarkably sensitive to interactions, since it assumes neither linear relationships, normal distributions, nor homoscedasticity. This programme is put to work here on a prediction problem of administrative importance: which women in Korean villages will adopt family planning under a mild programme and under an intensive programme? Optimum predictors are identified and are arranged in a hierarchy of combinations which give progressively higher predictive accuracy. The best two or three predictors isolate large proportions of women with extremely low adoption rates.

This publication has 3 references indexed in Scilit: