Generalizability diagnostics from trial and target sampling scores

For estimated trial-participation scores s(x) with logits l(x) = logit(s(x)), returns the standardized mean difference of the logits, \(SMD = (\bar l_{sample} - \bar l_{target}) / sd\), with sd the standard deviation of the logits in the target population (Tipton and Hartman 2023, Eq. 3.7, crediting Stuart et al. 2011, who propose the difference in mean sampling scores as the primary similarity metric; the raw-scale difference is reported as well). Positive values mean the sample sits higher on the participation-score scale than the target population.

Usage

Causgsw(s_sample, s_target)

Arguments

s_sample

Sampling scores of the trial sample, strictly in (0, 1).

s_target

Sampling scores of the target population, strictly in (0, 1).

Value

List with estimate (SMD of logits, sample minus target over the target logit sd), smd_abs, diff_means, mean_sample, mean_target, n_sample, n_target, method.

References

Tipton, E. and Hartman, E. (2023), Generalizability and transportability, Ch. 3 Eq. 3.7 in Zubizarreta, Stuart, Small and Rosenbaum (eds), Handbook of Matching and Weighting Adjustments for Causal Inference, Chapman and Hall/CRC, doi:10.1201/9781003102670 (local PDF, WD_BLACK library). Stuart, E. A., Cole, S. R., Bradshaw, C. P. and Leaf, P. J. (2011), The use of propensity scores to assess the generalizability of results from randomized trials, Journal of the Royal Statistical Society Series A 174(2), 369-386, doi:10.1111/j.1467-985x.2010.00673.x.

Examples

set.seed(1)
Causgsw(s_sample = runif(50, 0.1, 0.9), s_target = runif(30, 0.1, 0.9))
#> $estimate
#> [1] 0.04359192
#> 
#> $smd_abs
#> [1] 0.04359192
#> 
#> $diff_means
#> [1] 0.01469696
#> 
#> $mean_sample
#> [1] 0.5260743
#> 
#> $mean_target
#> [1] 0.5113773
#> 
#> $n_sample
#> [1] 50
#> 
#> $n_target
#> [1] 30
#> 
#> $method
#> [1] "Tipton-Hartman Eq. 3.7 SMD of sampling-score logits"
#>