Add spending time to sim_gs_n #340
Description
For group sequential designs, spending functions can be defined based on different fractions:
- information fraction (default)
- event fraction
- calendar time fraction
- minimal of planned and actual information fraction
The above 4 options is for interim analysis (IA). For final analysis (FA), there are 2 options:
- actual information fraction
- full alpha
To enable the above 8 options for a group sequential design, there are 2 proposed API.
API proposal 1:
sim_gs_n(
spending_time = c("info_frac", "event_frac", "time_frac", "min_planned_actual_info_frac"),
full_alpha_for_FA = TRUE,
....)
API proposal 2:
When specifying the spending time in sim_gs_n(), we may temporarily propose introducing a create_spending() function, similar to how we implemented create_cut(). For example:
create_spending_time <- function(actual_info = FALSE,
planned_info = FALSE,
actual_event = FALSE,
actual_time = FALSE){...}
Since the information fraction, event fraction, and calendar time fraction can vary between simulations, the spending function should be flexible to accommodate any combination of these inputs. The idea is to specify which of info, event, and time are used to define the spending.
Below is a draft illustrating this concept (please note the syntax and grammar are for illustration and may need refinement):
# spending by information fraction only (default)
# this is the same as API 1 when spending_time = "info_frac"
sim_gs_n(spending_time = create_spending_time(actual_info = TRUE),
full_alpha_for_FA = TRUE,
...)
# spending by event fraction only
# this is the same as API 1 when spending_time = "event_frac"
sim_gs_n(spending = create_spending_time(actual_event = TRUE)
full_alpha_for_FA = TRUE,
...)
# spending by calendar time fraction only
# this is the same as API 1 when spending_time = "time_frac"
sim_gs_n(spending = create_spending_time(actual_time = TRUE)
full_alpha_for_FA = TRUE,
...)
# spending by min{event fraction, calendar time fraction}
# this is the same as API 1 when spending_time = "min_planned_actual_info_frac"
sim_gs_n(spending = create_spending_time(actual_info = TRUE, planned_info = TRUE)
full_alpha_for_FA = TRUE,
...)
Comments and thoughts are welcome!!!