Checking assumptions in survival analysis with accelerated failure time

[Paul T Seed]

When fitting a proportonal hazards model, it is common enough to plot the (log) hazard funciton against time for different groups of subjects, and check that the lines are parallel; thus confirming the proportional hazards assumption. The same can be done when fitting a parametric model (using streg) that has the proportonal hazards assumption.,

When we fit models of the accelerated time type, such as loglogistic and lognormal, is there any similar plot to check the assumptions? The only thing I can find is the Cox-Snell plot in example 4 of stcox postestimation.

[Ronald McDowell]

I also am interested in this question so am reposting. My understanding is that the underlying assumption is that the effect of covariates act multiplicatively (proportionally) with respect to the survival time.There isn't any command syntax in which this can be readily checked in Stata. I would be interested to know how others have gone about this, or if there are papers which discuss how to undertake the necessary steps.