Subclassification (stratification) on the propensity score

Native propensity-score subclassification with the rules of MatchIt::matchit(method = "subclass", estimand = "ATT"): the cut points are quantiles of the treated units' scores, every unit falls in the subclass of its score, and a subclass left without a treated or a control unit takes the nearest unit of that group from the closest subclass that can spare one ("scooting", MatchIt's min.n = 1). Weights are 1 for treated units and the treated/control ratio of the subclass for controls, each group rescaled to its size. Cross-validated against MatchIt in tests/cross. No MatchIt at runtime.

Usage

morie_matching_subclassify(
  data,
  treatment,
  covariates,
  ps = NULL,
  n_strata = 5L
)

Arguments

data

Data frame.

treatment

Binary treatment column name.

covariates

Character vector of covariates.

ps

Optional propensity scores (as for morie_matching_full()).

n_strata

Number of quantile-based strata (default 5).

Value

A list with components data_with_strata (the data with distance, weights, subclass, ._ps and ._stratum columns) and stratum_effects (per-stratum sample sizes and propensity-score ranges).

Examples

set.seed(1)
df <- data.frame(y = rnorm(200), d = rbinom(200, 1, 0.4),
                 x1 = rnorm(200), x2 = rnorm(200))
morie_matching_subclassify(df, "d", c("x1", "x2"), n_strata = 5)$stratum_effects
#>   stratum n_treated n_control ps_range_low ps_range_high
#> 1       1        18        27    0.2956479     0.3927630
#> 2       2        17        24    0.3934888     0.4238873
#> 3       3        17        24    0.4244656     0.4584856
#> 4       4        17        23    0.4596167     0.4877596
#> 5       5        18        15    0.4894480     0.5805204