A large-sample confidence interval for the inverse prediction of quantile differences in logistic regression for two independent tests

摘要

In the field of flight test, logistic regression (which models a dichotomous response variable as a function of covariates) has wide applicability. It is common to record a response as {hit, miss} and to count the number of hits (successes) at each level of input, so that response is a quantal variable. The differences in range due to possibly different radar equipment configurations are to be measured in the detection performance. This article develops an analytical approach to derive a symmetric confidence interval approximation for the average difference, which needs no simulation. The results are based on large-sample properties of ML estimates and this effort extends an existing result in nonlinear modeling (Ref. 5). The proposed confidence interval ensures good coverage probabilities as demonstrated through simulation results.