Propensity-score overlap diagnostics

Reports the common-support interval shared by the treated and control propensity scores, the percentage of observations that trimming to that interval would discard (Crump et al. 2009), and a two-sample Kolmogorov-Smirnov comparison of the two score distributions. Overlap is the positivity half of the strong ignorability condition of Rosenbaum and Rubin (1983); without it no matching or weighting estimator is identified in the affected region.

Usage

morie_over(ps_treated, ps_control)

Arguments

ps_treated

Propensity scores of treated units.

ps_control

Propensity scores of control units.

Value

A list with overlap_range (c(0, 0) when the supports do not intersect), pct_trimmed, ks_stat, ks_p, n_treated and n_control.

References

Crump, R. K., Hotz, V. J., Imbens, G. W. and Mitnik, O. A. (2009). Dealing with limited overlap in estimation of average treatment effects. Biometrika, 96(1), 187-199.

Examples

V <- c(1, 2, 3, 4, 5, 6, 7, 8)
morie_over(V, V)
#> $overlap_range
#> [1] 1 8
#> 
#> $pct_trimmed
#> [1] 0
#> 
#> $ks_stat
#> [1] 0
#> 
#> $ks_p
#> [1] 1
#> 
#> $n_treated
#> [1] 8
#> 
#> $n_control
#> [1] 8
#>