Decimals display in summary for a group sequential subject

[LittleBeannie]

The following lines of codes gives me an error message:

library(gsDesign2)

design <- gs_design_ahr(
  # time of interim and final analyses
  analysis_time = c(12, 24),
  # spending bound for efficacy
  upper = gs_spending_bound, 
  # O’Brien-Fleming spending function; total_spend is normally alpha
  upar = list(sf = gsDesign::sfLDOF, total_spend = 0.025),
  # fixed Z-values will be provided for futility bound
  lower = gs_b, 
  lpar = c(qnorm(0.1), -Inf, -Inf)
)

design |>
  summary(
    col_vars = c(
      "z", "~hr at bound", "nominal p", "Alternate hypothesis", "Null hypothesis"
    ),
    col_decimals = c(4, 4, 4, 4, 4)
  ) |>
    as_gt()

[jdblischak]

I confirmed I can reproduce the bug:

git2r::commits()[[1]]
## [4a97d17] 2024-06-13: Merge pull request #420 from Merck/419-add-developer-test-for-gs_power_ahr
library(gsDesign2)

design <- gs_design_ahr(
  # time of interim and final analyses
  analysis_time = c(12, 24),
  # spending bound for efficacy
  upper = gs_spending_bound,
  # O’Brien-Fleming spending function; total_spend is normally alpha
  upar = list(sf = gsDesign::sfLDOF, total_spend = 0.025),
  # fixed Z-values will be provided for futility bound
  lower = gs_b,
  lpar = c(qnorm(0.1), -Inf, -Inf)
)

design |>
  summary(
    col_vars = c(
      "z", "~hr at bound", "nominal p", "Alternate hypothesis", "Null hypothesis"
    ),
    col_decimals = c(4, 4, 4, 4, 4)
  ) |>
  as_gt()
## Adding missing grouping variables: `Analysis`
## Error in `dplyr::filter()`:
## ℹ In argument: `Bound == display_bound[1]`.
## ℹ In group 1: `Analysis = "Analysis: 1 Time: 12 N: 701.7 Event: 159.1 AHR: 0.81 Information
##   fraction: 0.42"`.
## Caused by error:
## ! object 'Bound' not found
## Run `rlang::last_trace()` to see where the error occurred.

sessionInfo()
## R version 4.3.3 (2024-02-29 ucrt)
## Platform: x86_64-w64-mingw32/x64 (64-bit)
## Running under: Windows 11 x64 (build 22631)
##
## Matrix products: default
##
##
## locale:
## [1] LC_COLLATE=English_United States.utf8  LC_CTYPE=English_United States.utf8
## [3] LC_MONETARY=English_United States.utf8 LC_NUMERIC=C
## [5] LC_TIME=English_United States.utf8
##
## time zone: America/New_York
## tzcode source: internal
##
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base
##
## other attached packages:
## [1] gsDesign2_1.1.2.11
##
## loaded via a namespace (and not attached):
##  [1] Matrix_1.6-5      gtable_0.3.5      dplyr_1.1.4       compiler_4.3.3    tidyselect_1.2.1
##  [6] Rcpp_1.0.12       xml2_1.3.6        tidyr_1.3.1       splines_4.3.3     scales_1.3.0
## [11] fastmap_1.1.1     lattice_0.22-6    ggplot2_3.5.1     R6_2.5.1          generics_0.1.3
## [16] tibble_3.2.1      r2rtf_1.1.1       munsell_0.5.1     pillar_1.9.0      rlang_1.1.3
## [21] utf8_1.2.4        cli_3.6.2         withr_3.0.0       magrittr_2.0.3    digest_0.6.35
## [26] grid_4.3.3        rstudioapi_0.16.0 mvtnorm_1.2-4     xtable_1.8-4      gsDesign_3.6.2
## [31] lifecycle_1.0.4   vctrs_0.6.5       glue_1.7.0        data.table_1.15.4 survival_3.6-4
## [36] gt_0.10.1         fansi_1.0.6       colorspace_2.1-0  purrr_1.0.2       tools_4.3.3
## [41] pkgconfig_2.0.3   htmltools_0.5.8.1

[jdblischak]

From the backtrace, the error occurs in as_gt.gs_design()

gsDesign2/R/as_gt.R

Line 309 in 4a97d17

x_alpha <- max((x %>% dplyr::filter(Bound == display_bound[1]))[[colname_spannersub[2]]])

rlang::last_trace()
## <error/rlang_error>
## Error in `dplyr::filter()`:
## ℹ In argument: `Bound == display_bound[1]`.
## ℹ In group 1: `Analysis = "Analysis: 1 Time: 12 N: 701.7 Event: 159.1 AHR: 0.81 Information
##   fraction: 0.42"`.
## Caused by error:
## ! object 'Bound' not found
## ---
## Backtrace:
##      ▆
##   1. ├─base::source("~/.active-rstudio-document", echo = TRUE)
##   2. │ ├─base::withVisible(eval(ei, envir))
##   3. │ └─base::eval(ei, envir)
##   4. │   └─base::eval(ei, envir)
##   5. ├─gsDesign2::as_gt(...) at ~/.active-rstudio-document:15:1
##   6. ├─gsDesign2:::as_gt.gs_design(...) at gsDesign2/R/as_gt.R:28:3
##   7. │ └─x %>% dplyr::filter(Bound == display_bound[1]) at gsDesign2/R/as_gt.R:309:3
##   8. ├─dplyr::filter(., Bound == display_bound[1])
##   9. └─dplyr:::filter.data.frame(., Bound == display_bound[1])
##  10.   └─dplyr:::filter_rows(.data, dots, by)
##  11.     └─dplyr:::filter_eval(...)
##  12.       ├─base::withCallingHandlers(...)
##  13.       └─mask$eval_all_filter(dots, env_filter)
##  14.         └─dplyr (local) eval()
## Run rlang::last_trace(drop = FALSE) to see 3 hidden frames.

[jdblischak]

I was worried that I introduced the bug in #403, but it appears to have existed prior to this. Below I reproduced the bug from commit 9071ab1

git2r::commits()[[1]]
## [9071ab1] 2024-05-23: Merge pull request #399 from Merck/397-update-gs_update_ahr-to-get-rid-of-ia_alpha_spending-and-fa_alpha_spending
library(gsDesign2)
packageVersion("gsDesign2")
## [1] ‘1.1.2.7’

design <- gs_design_ahr(
  # time of interim and final analyses
  analysis_time = c(12, 24),
  # spending bound for efficacy
  upper = gs_spending_bound,
  # O’Brien-Fleming spending function; total_spend is normally alpha
  upar = list(sf = gsDesign::sfLDOF, total_spend = 0.025),
  # fixed Z-values will be provided for futility bound
  lower = gs_b,
  lpar = c(qnorm(0.1), -Inf, -Inf)
)

design |>
  summary(
    col_vars = c(
      "z", "~hr at bound", "nominal p", "Alternate hypothesis", "Null hypothesis"
    ),
    col_decimals = c(4, 4, 4, 4, 4)
  ) |>
  as_gt()
## Adding missing grouping variables: `Analysis`
## Error in `dplyr::filter()`:
## ℹ In argument: `Bound == display_bound[1]`.
## ℹ In group 1: `Analysis = "Analysis: 1 Time: 12 N: 701.7 Event: 159.1 AHR: 0.81 Information
##   fraction: 0.42"`.
## Caused by error:
## ! object 'Bound' not found
## Run `rlang::last_trace()` to see where the error occurred.

[jdblischak]

I don't think this bug is a regression. I think this is how it has always behaved. You have to include "bound" in the vector passed to col_vars.

I couldn't find any tests or examples that use col_vars, but the article story-update-boundary.Rmd uses col_vars mutiple times, and it always includes "bound", eg

gsDesign2/vignettes/articles/story-update-boundary.Rmd

Lines 134 to 140 in 4a97d17

	summary(
	col_vars = c(
	"analysis", "bound", "z", "~hr at bound",
	"nominal p", "Alternate hypothesis", "Null hypothesis"
	),
	col_decimals = c(NA, NA, 4, 4, 4, 4, 4)
	) |>

If I add "bound" to your example, it runs:

library(gsDesign2)

design <- gs_design_ahr(
  # time of interim and final analyses
  analysis_time = c(12, 24),
  # spending bound for efficacy
  upper = gs_spending_bound,
  # O’Brien-Fleming spending function; total_spend is normally alpha
  upar = list(sf = gsDesign::sfLDOF, total_spend = 0.025),
  # fixed Z-values will be provided for futility bound
  lower = gs_b,
  lpar = c(qnorm(0.1), -Inf, -Inf)
)

design |>
  summary(
    col_vars = c(
      "bound", "z", "~hr at bound", "nominal p", "Alternate hypothesis", "Null hypothesis"
    ),
    col_decimals = c(NA, 4, 4, 4, 4, 4)
  ) |>
  as_gt()
## Adding missing grouping variables: `Analysis`

Thus I think this warrants a discussion around the argument col_vars. Should "bound" always be included? And if yes, then should we update the behavior so that the user doesn't have to manually specify it?

[LittleBeannie]

I don't think this bug is a regression. I think this is how it has always behaved. You have to include "bound" in the vector passed to col_vars.

I couldn't find any tests or examples that use col_vars, but the article story-update-boundary.Rmd uses col_vars mutiple times, and it always includes "bound", eg

gsDesign2/vignettes/articles/story-update-boundary.Rmd

Lines 134 to 140 in 4a97d17

	summary(
	col_vars = c(
	"analysis", "bound", "z", "~hr at bound",
	"nominal p", "Alternate hypothesis", "Null hypothesis"
	),
	col_decimals = c(NA, NA, 4, 4, 4, 4, 4)
	) |>

If I add "bound" to your example, it runs:

library(gsDesign2)

design <- gs_design_ahr(
  # time of interim and final analyses
  analysis_time = c(12, 24),
  # spending bound for efficacy
  upper = gs_spending_bound,
  # O’Brien-Fleming spending function; total_spend is normally alpha
  upar = list(sf = gsDesign::sfLDOF, total_spend = 0.025),
  # fixed Z-values will be provided for futility bound
  lower = gs_b,
  lpar = c(qnorm(0.1), -Inf, -Inf)
)

design |>
  summary(
    col_vars = c(
      "bound", "z", "~hr at bound", "nominal p", "Alternate hypothesis", "Null hypothesis"
    ),
    col_decimals = c(NA, 4, 4, 4, 4, 4)
  ) |>
  as_gt()
## Adding missing grouping variables: `Analysis`

Thus I think this warrants a discussion around the argument col_vars. Should "bound" always be included? And if yes, then should we update the behavior so that the user doesn't have to manually specify it?

Thanks for researching this!

• "bound" is always a column to be displayed in the table, so people can tell whether it is an efficacy test or futility test.
• However, when changing the decimals of columns, we do not need to specify the decimals of "bound", because it is a character (either bound = "upper" or bound = "lower").