Bootstrap results (SEs and CIs) not appearing despite n_boot and boot_diag enabled

[tanyatoluay]

Issue:
I'm running the gformula() function with bootstrapping enabled (n_boot = 1000) and diagnostics (boot_diag = TRUE) to obtain standard errors and confidence intervals. However, the output still contains NA for SEs and CIs, and calling summary(gform_result)$boot_table returns NULL.

The model itself seems to fit well and produces predicted means, but no bootstrap results are saved or shown.

What I've tried:

• Set n_boot = 1000 in the function call.
• Enabled boot_diag = TRUE to get diagnostic output.
• Verified the model fits without warnings or errors.
• Attempted to inspect bootstrapping diagnostics (no clear failures shown).
• Checked for bootstrap results in the returned object (boot_table, boot_diag_results) but they are missing or NULL.
• Made sure to set a random seed.

Minimal reproducible example code:

gform_result <- gformula(
  obs_data = df_final,
  outcome_type = outcome_type,
  id = id_name,
  time_name = time_name,
  time_points = time_points,
  outcome_name = outcome_name,
  covnames = covnames,
  basecovs = basecovs,
  covtypes = covtypes,
  histories = histories,
  histvars = histvars,
  covparams = covparams,
  ymodel = ymodel,
  intvars = intvars,
  interventions = interventions,
  int_descript = int_descript,
  nsimul = nsimul,
  n_boot = 1000,
  seed = seed,
  parallel = parallel,
  nsamples = nsamples,
  model_fits = TRUE,
  boot_diag = TRUE
)

summary(gform_result)$boot_table  # returns NULL

Question:
Is there something else I need to enable or check to get bootstrap SEs and confidence intervals in the output? Are there known issues with bootstrapping results not being saved?

[stmcg]

Hi,

Thanks for posting the question! One issue is that the bootstrap results aren't saved in summary(gform_result)$boot_table. They should be displayed when printing the object gform_result if bootstrapping was performed.

My first suggestion would be to review the help file of the gformula function and revise the argument names in your code so that they are all specified correctly. For example, the gformula function doesn't take the argument n_boot. Instead, the number of bootstrap samples is set by the nsamples argument, which may be why you aren't finding bootstrap being performed.

If you're still finding the error, could you try using these settings on one of the data sets in the package and share the code? That'd be very helpful to sort this out.

As for your last question, there aren't any known issues with bootstrapping results not being saved.

Some small tips: For the sake of figuring out if the package is performing bootstrapping, I'd suggest setting nsamples to something small (e.g., 5) to reduce computation time. It'd might also help to simplify your call to the gformula function, such as by removing parallelization.

Best,
Sean