Because I have been preparing an extended presentation on (Bayesian) survival analysis in health economic evaluation, I took the opportunity to make some tweaks to survHE — nothing major, but I was aware of couple of imprecisions in the code or things I wanted to make a bit better, so while I was knitring my slides, I made the changes.
As soon as I get a moment, I’ll also update my repository (which can be used to update the package), but for now, I’ve uploaded the changes on GitHub, so the development version (1.0.63) is now available by simply typing
on the R terminal (assuming you have installed survHE).
The changes are the following:
I’ve fixed a minor problem in the print method; up until now, if you were fitting an Exponential model with no covariates using MLE (ie via flexsurv), then the resulting output from the call to print was simply showing the values (mean, 95% confidence interval and SE) but no label for the relevant parameter (it should have read rate). The print method was doing OK when the underlying model was run using HMC (via rstan) or INLA. Now it consistently report the name of the parameters.
I’ve made some addition to the model.fit.plot, which can be used to visualise the model fitting using Information Criteria, as recommended by the NICE DSU. Now, there’s an additional option scale, which by default is set to “absolute” (and nothing changes in comparison to the previous version) but can now also take on the value “relative” (or “rel”). This in turn creates barplots with the percentage increase in the *IC selected with comparison to the best fitting model (among the lot selected). This makes it easier to check how much better the “better” model is.
I may make a few more changes in the next few days, as I keep working on this — and if so, I’ll update the GitHub repository.