PSsurvival (version 0.2.0)

estimate_weights: Estimate Propensity Score Weights

Description

Calculates propensity score weights for causal inference with optional trimming. Supports ATE, ATT, and overlap population estimands for binary and multiple treatment groups.

Usage

estimate_weights(
  ps_result,
  data,
  treatment_var,
  estimand = "ATE",
  att_group = NULL,
  trim = NULL,
  delta = NULL,
  alpha = NULL
)

Value

A list containing:

weights

Numeric vector of weights (length = nrow(data)).

trim_summary

Data frame with trimming summary by treatment group.

ess

Named numeric vector of effective sample size by treatment group.

method

Character string: "IPW" for ATE/ATT, "overlap" for overlap.

estimand

Character string of estimand used.

att_group

Target group for ATT (NULL if not applicable).

trim_method

Character string of trimming method (NULL if no trimming).

delta

Numeric trimming threshold used for symmetric trimming (NULL if not applicable).

alpha

Numeric percentile threshold used for asymmetric trimming (NULL if not applicable).

n_levels

Number of treatment levels.

treatment_levels

Vector of treatment level values.

ps_result

PS result object (refitted after trimming if trimming was applied).

Arguments

ps_result

A list returned by estimate_ps(), containing the fitted propensity score model and estimated propensity scores.

data

A data.frame containing the treatment variable (same data used in estimate_ps()).

treatment_var

A character string specifying the name of the treatment variable in data.

estimand

Character string specifying the target population. One of:

• "ATE": Average Treatment Effect (default). Uses IPW method.

• "ATT": Average Treatment Effect on the Treated. Uses IPW method.

• "overlap": Overlap population (Li & Li, 2019). Uses overlap weighting.

att_group

For ATT estimation, specifies which treatment group to target. This is MANDATORY when estimand = "ATT". Ignored for other estimands.

trim

Character string specifying the trimming method, or NULL for no trimming (default). Options: "symmetric" (Crump extension) or "asymmetric" (Sturmer extension). Trimming is NOT supported with overlap estimand.

delta

Trimming threshold for symmetric trimming in (0, 1/J], where J is the number of treatment levels. If NULL (default), uses recommended values from Yoshida et al. (2019). Ignored unless trim = "symmetric".

alpha

Percentile threshold for asymmetric trimming in (0, 0.5). If NULL (default), uses recommended values from Yoshida et al. (2019). Ignored unless trim = "asymmetric".

Details

Trimming Workflow: When trimming is requested, the function: (1) identifies observations to trim using PS from full data, (2) re-estimates PS on trimmed data, (3) calculates weights from re-estimated PS. This ensures trimming uses the original covariate distribution while weights reflect the overlapping population.

Overlap weights do not support trimming (already bounded in [0,1]).

References

Li, F., & Li, F. (2019). Propensity score weighting for causal inference with multiple treatments. The Annals of Applied Statistics, 13(4), 2389-2415.

Yoshida, K., et al. (2019). Multinomial extension of propensity score trimming methods: A simulation study. American Journal of Epidemiology, 188(3), 609-616.

Crump, R. K., et al. (2009). Dealing with limited overlap in estimation of average treatment effects. Biometrika, 96(1), 187-199.

Examples

# \donttest{
# Example 1: Overlap weighting for binary treatment
data(simdata_bin)
ps_bin <- estimate_ps(
  data = simdata_bin,
  treatment_var = "Z",
  ps_formula = Z ~ X1 + X2 + X3 + B1 + B2
)
weights_ow <- estimate_weights(
  ps_result = ps_bin,
  data = simdata_bin,
  treatment_var = "Z",
  estimand = "overlap"
)
summary(weights_ow$weights)

# Example 2: ATT with multiple treatments
data(simdata_multi)
ps_multi <- estimate_ps(
  data = simdata_multi,
  treatment_var = "Z",
  ps_formula = Z ~ X1 + X2 + X3 + B1 + B2
)
weights_att <- estimate_weights(
  ps_result = ps_multi,
  data = simdata_multi,
  treatment_var = "Z",
  estimand = "ATT",
  att_group = "C"
)
summary(weights_att$weights)
# }