Wrong comparison of intervals in warn_about_new_time_points.pamm?

[gavinsimpson]

I think there's a little buglet in the code to compare intervals used in a PAMM and in newdata supplied for prediction. In warn_about_new_time_points.pamm we have:

pammtools/R/warnings.R

Line 44 in ab4caea

if(!all(int_new %in% int_original)) {

In the above, the code is comparing int_new which is the result of

pammtools/R/warnings.R

Line 43 in ab4caea

int_new <- unique(newdata[["interval"]])

which has extracted the unique event times tend in newdata. But it's comparing with the output from int_info(object), computed in L41

pammtools/R/warnings.R

Line 41 in ab4caea

int_original <- int_info(object)

which is a data frame.

Shouldn't line 44 therefore be

if(!all(int_new %in% int_original[["interval"]])) {

to compare the interval column used in the model with the ones supplied in newdata?

[gavinsimpson]

I see where this happened; it got changed in 1bf8b74

I'll pull a quick PR together.

[adibender]

Thanks Gavin, will fix/merge soon.
I was thinking of removing the warning anyway. Once you estimated the baselinehazard, I don't see a reason to then not allow predictions at any time point (or use a finer/different grid compared to the grid used at estimation stage), as we can evaluate the smooth function at any time point.

[adibender]

closed in #236