Hypothesis Tests for Limited Failure Survival Distributions

Publisher Website
Google Scholar
Abstract
A family of survival distributions that are mixtures of one group which fails according to an exponential law with another group that does not fail has recently come into use in correctional program evaluation. The concept is generalized to include mixtures based on the Weibull distribution. Available graphical and numeric goodness-of-fit tests are discussed and a new test for grouped, multiply censored data due to Turnbull and Weiss (1978) is used to discriminate between the various special cases in this family of mixtures. The goodness-of-fit test is also extended to allow k group tests of hypotheses on the parameters.