Matt Warkentin
Matt Warkentin
Okay @hfrick, we are getting close. The one thing I'm stumbling on is that there surely needs to be somewhere where we connect the fact that OUR parameter is called...
> Could you update them so that they make use of the spline functionality? Then we know that this aspect also works! For survival probability and hazard, we don't need...
> For flexsurvreg(), predictions by flexsurv are exp(x * beta), which is why censored logs them before returning them as predictions of type linear_pred (so that it's x * beta)....
@hfrick You may wish to use the functions in `x$dlist$inv.transforms` to get location parameters on the unrestricted scale. As Chris mentioned, it is often either `identity()` or `log()`.
Oops, my mistake. Please ignore my previous comment.
I see you using the `summary.flexsurvreg` method for predictions (both here and in the package source code). You may wish to use `predict.flexsurvreg` which I've implemented to be very tidy-friendly....
I'm not sure what you mean. A PR for this package or a PR for `{flexsurv}`? The `predict.flexsurvreg` method has been available in the dev version since 2020-06-10 and on...
As always, happy to be a contributor and not just a beggar.
Could also just add `'spline'` as a distribution in `dials::values_flexsurv_dist` and add the argument `k` to `parsnip::survival_reg()`: ```r survival_reg(engine = 'flexsurv', dist = 'spline', k = tune()) # or knots...
I think you are right and it makes sense for this to be implemented as a separate engine. I am working on updating the `dials` PR as we speak. For...