Two-way analysis of reciprocals

Abstract

SUMMARY A failure-time model, which leads to inverse Gaussian random samples, with common ratios of mean to variance, is considered. Hypothesis testing problems, analogous to those arising in two-way analysis of variance, are discussed in detail. The work of Tweedie (1957) plays an important role in the development, despite the fact that his analysis of reciprocals procedure has a different homogeneity assumption and does not deal with crossed classifications, and hence does not directly apply to the model under investigation.