Skip to contents

Calculate results across binary-imputation scenarios

Source: R/resp_imputation.R
resp_multiple_imputation.Rd

Runs h_impute_analyze_resp() for each paired control and treatment response probability, pools estimates across imputations using Rubin's rules, and pools Wilson-Hilferty z statistics to obtain a p-value. A scenario where both probabilities are zero or one is deterministic and is therefore analyzed once rather than repeatedly imputed.

Usage

resp_multiple_imputation(
  dat,
  p_ctrl,
  p_trt,
  trtvar,
  ctrlab,
  trtlab,
  respvar,
  stratvar,
  n_imputations,
  pvalcat
)

Arguments

dat

(data.frame) Analysis data containing treatment, response, and stratification variables.

p_ctrl

(numeric) Control-group response probabilities, one for each scenario.

p_trt

(numeric) Treatment-group response probabilities, paired with p_ctrl by position.

trtvar

(string) Name of the treatment variable in dat.

ctrlab

(string) Value of trtvar identifying the control group.

trtlab

(string) Value of trtvar identifying the treatment group.

respvar

(string) Name of a logical binary response variable in dat.

stratvar

(character) Names of one or more stratification variables in dat.

n_imputations

(count) Number of imputations for non-deterministic scenarios.

pvalcat

(named list) P-value categories passed to categorize_pval().

Value

A tibble with one row per scenario. rd and rd_se are the pooled risk-difference estimate and standard error; p_value and p_cat are the pooled p-value and its category. marker_flag and effect_label contain "*" when a deterministic scenario or a single-imputation p-value fallback was used.

Examples

dat <- data.frame(
  arm = rep(c("Control", "Treatment"), each = 4),
  response = c(TRUE, FALSE, NA, TRUE, TRUE, FALSE, NA, FALSE),
  strata = rep(c("A", "B"), 4)
)
pvalcat <- list("<0.05" = c(0, 0.05), ">=0.05" = c(0.05, 1))

set.seed(123)
resp_multiple_imputation(
  dat,
  p_ctrl = c(0, 0.25),
  p_trt = c(1, 0.75),
  trtvar = "arm",
  ctrlab = "Control",
  trtlab = "Treatment",
  respvar = "response",
  stratvar = "strata",
  n_imputations = 5,
  pvalcat = pvalcat
)
#> Warning: Less than 5 observations in some strata.
#> Warning: <5 data points in some strata. CMH test may be incorrect.
#> Warning: Less than 5 observations in some strata.
#> Warning: <5 data points in some strata. CMH test may be incorrect.
#> Warning: Less than 5 observations in some strata.
#> Warning: <5 data points in some strata. CMH test may be incorrect.
#> Warning: Less than 5 observations in some strata.
#> Warning: <5 data points in some strata. CMH test may be incorrect.
#> Warning: Less than 5 observations in some strata.
#> Warning: <5 data points in some strata. CMH test may be incorrect.
#> Warning: Less than 5 observations in some strata.
#> Warning: <5 data points in some strata. CMH test may be incorrect.
#> # A tibble: 2 × 9
#>   p_ctrl p_trt     m    rd rd_se p_value p_cat  marker_flag effect_label
#>    <dbl> <dbl> <dbl> <dbl> <dbl>   <dbl> <chr>  <chr>       <chr>       
#> 1   0     1        1   0    0.25   1     >=0.05 "*"         0.0%*       
#> 2   0.25  0.75     5  -0.2  0.3    0.951 >=0.05 ""          -20.0%