Matt Warkentin

Thanks for all the information. I have some ideas for how to perhaps fix this. Working on it now and will report back soon.

Thanks for the quick reply. Yes, perhaps `bibtex` would be a good long-term solution if it is actively being maintained now. So long as it meets the same needs as...

> Do we need a sanitize_model_specific() method to return an informative error if the user does not supply times? We may. I made a design consideration to force the requirement...

> A couple simple tests if possible. Do you know if we can compare to benchmark estimates from alternative software? We may be able to benchmark against the `stpm2` implementation...

I agree that users will likely want to predict at several time horizons. Glad to see this type of multi-prediction is already supported by this package. This would actually allow...

The list-columns are predictable, and I have now solved that issue. I think I may need to add some info to `type_dictionary_build()` to allow support the various prediction types. Right?...

I get this warning when trying to pass the `times` argument into `predictions()`: >Warning message: These arguments are not supported for models of class `flexsurvreg`: times. Please file a request...

I've made those changes. Getting closer. After adding valid arguments to `sanity_dots.R`, I am now running into the following issue with what seems like partial arg matching between the `p`...

Yeah, bit of a tricky one. I will keep thinking it over. If it becomes too much of hassle we can drop the `quantile` prediction type (and thus the need...

Sorry, I was away camping for the long weekend. Will get back to this shortly.