`check_model` for `rms:lrm`

There is some minimal support for rms::lrm already, for example:

model = rms::lrm(formula=am ~ mpg, data=mtcars, x=T, y=T, model=T)
print(performance::model_performance(model))

returns

Can't calculate log-loss.
Indices of model performance

R2	RMSE	Sigma
0.466	0.392	0.995

This is rudimentary compared to the base logistic regression support:

model_base = stats::glm(formula=am ~ mpg, data=mtcars, family='binomial')
print(performance::model_performance(model_base))

Indices of model performance

AIC	BIC	Tjur's R2	RMSE	Sigma	Log_loss	Score_log	Score_spherical	PCP
33.675	36.607	0.368	0.392	0.995	0.464	-9.303	0.093	0.695

And check_model(model) does not work at all:

Error in if (model$rank == 0) { : argument is of length zero
     ▆
  1. ├─base::withVisible(...)
  2. ├─performance::check_model(model)
  3. ├─performance:::check_model.default(model)
  4. │ ├─base::suppressWarnings(.check_assumptions_glm(x, minfo))
  5. │ │ └─base::withCallingHandlers(...)
  6. │ └─performance:::.check_assumptions_glm(x, minfo)
  7. │   └─performance:::.diag_qq(model)
  8. │     ├─base::sort(stats::rstandard(model, type = "pearson"), na.last = NA)
  9. │     ├─stats::rstandard(model, type = "pearson")
 10. │     └─stats:::rstandard.glm(model, type = "pearson")
 11. │       ├─stats::influence(model, do.coef = FALSE)
 12. │       └─stats:::influence.glm(model, do.coef = FALSE)
 13. │         └─stats::lm.influence(model, do.coef = do.coef, ...)
 14. └─global `<fn>`()

Error in if (model$rank == 0) { : argument is of length zero

So my questions is: do you plan to support rms::lrm, or is it not worth it since most common use cases can be handled with stats::glm?

Start with check_model(model) and the check_model.default path shown in the traceback, especially .check_assumptions_glm and .diag_qq. Compare the existing rms::lrm behavior with stats::glm and model_performance(model), then verify that check_model works and the missing performance measures are supported without the reported error.