What are the multiple correlations of three sets of predictors and overall state of health? The first set of predictors contains demographic variables (age and years of education). The second….
compare a 5-point discrete mixture on the log-logistic shape parameter with the variable scale model to downweight aberrant cases, namely ui ∼ L(ηi, 1/(κθi)) where θi are gamma with mean 1, and ui = log(ti).
In Example 13.2 compare a 5-point discrete mixture on the log-logistic shape parameter with the variable scale model to downweight aberrant cases, namely ui ∼ L(ηi, 1/(κθi)) where θi are gamma with mean 1, and ui = log(ti).
Commuter delay in work-to-home trips Washington et al. (2003) consider the durations of delay in work-to-home trips for 96 Seattle area commuters. For such workers, the home trip is postponed to varying degrees to avoid evening rush-hour congestion. The hazard rate is effectively modelling early as against late departures for home. There is no censoring. The predictors are gender, X1(M = 1, F = 0), X2 = ratio of actual travel time (at expected departure time) to free-flow travel time, X3 = distance from work to home (km) and X4 = resident population density in workplace zone (divided by 10 000). One might expect early departure to be negatively associated with X2 and X4. Actual delay times vary from 4 to 240 min. A non-monotonic hazard is suggested when the Kaplan–Meier survival curve is used to provide estimates of H(t) and hence h(t). The hazard is low at first (durations under 20 min), has a plateau at values of 0.017 to 0.031 per minute for durations between 20 and 100 min and has a late peak between 120 and 140 min. Here we compare a Weibull with single-component and two-component log-logistic models.