Propensity-score subclassification ATE on OTIS data (native)

Rosenbaum-Rubin (1983) propensity-score stratification with the Cochran (1968) five-stratum convention: PS quintile strata, within-stratum treated-control mean differences, aggregated with stratum-size weights; SE from the weighted within-stratum variances. Strata missing either arm are dropped (and reported).

Usage

morie_otis_psm_subclass(df, treatment, outcome, covariates, n_strata = 5L)

Arguments

df

Data frame.

treatment, outcome, covariates

Column names.

n_strata

Number of PS strata. Default 5.

Value

An OTIS causal-estimate list (ATE).

Examples

set.seed(1)
n <- 400
x <- rnorm(n)
d <- rbinom(n, 1, plogis(0.6 * x))
y <- 1.5 * d + x + rnorm(n)
morie_otis_psm_subclass(data.frame(y, d, x), treatment = "d",
                        outcome = "y", covariates = "x")
#> $estimator
#> [1] "PS-subclass"
#> 
#> $ate
#> [1] 1.657085
#> 
#> $ate_se
#> [1] 0.125336
#> 
#> $ate_pval
#> [1] 6.625251e-40
#> 
#> $ate_ci95
#> [1] 1.411426 1.902744
#> 
#> $n
#> [1] 400
#> 
#> $n_treated
#> [1] 194
#> 
#> $p_treat
#> [1] 0.485
#> 
#> $notes
#> $notes[[1]]
#> [1] "strata=5 (dropped 0 without both arms)"
#> 
#> 
#> $strata
#>      stratum  n     diff
#> [1,]       1 80 1.960552
#> [2,]       2 80 1.797201
#> [3,]       3 80 1.231225
#> [4,]       4 80 1.597197
#> [5,]       5 80 1.699250
#> 
#> attr(,"class")
#> [1] "morie_causal_estimate" "list"