Extracting model parameters from INLA survival models

[brant...@gmail.com]

When an analyst fits a parametric survival model to a dataset, using MLE, INLA, or whatever other method, there are some key "returns" from the function that should be *easily* accessible to the analyst. These are the bits of information that characterize the model and that the analyst absolutely needs (e.g. parameters for the distribution and their SE, model goodness of fit statistics (AIC, BIC, DIC, etc...).   For example, if we fit an exponential model to a set of censored data, one might expect that the "rate" parameter would be easily and obvious to extract from the model fit. If a Weibull was fit, then the shape and scale parameters should be easy to see and utilize.

One problem I find with INLA is that it is very hard to find these critical model parameters in one spot. For example, one might expect a command like coef(fit.inla) would return these parameters. Is there a command that I am unaware of that can return the expected parameters from Weibull, exponential, log-logistic, and log-normal models fit in INLA?

Brant

[Janet Van Niekerk]

Hi Brant,

It is easy to extract the parameters using summary().

The hyperparameters (shape of Weibull) is also in summary.hyperpar().

Chapter 10 of Virgilio's book shows exactly how to get and use all these parameters.