Companion R package to the methodological paper:
Nakamura, D. (in press). The Denominator Chooses the Estimand: A Target-Population True-Score Framework for Standardized Mean Differences. Psychological Methods.
# From CRAN:
install.packages("ltgsmd")
# Development version from OSF:
# https://doi.org/10.17605/OSF.IO/KW9R6library(ltgsmd)
# Single study with an internal holdout reference
result <- compute_ltg_smd(
study_data = study_df,
reference_data = reference_df,
group_var = "condition",
score_var = "outcome",
items = c("item1", "item2", "item3", "item4"),
group_levels = c(reference = "control", focal = "treatment")
)
print(result)
# Confidence intervals
ci <- ltg_smd_ci(result, method = c("analytic", "bootstrap"),
B = 2000, seed = 20240501,
study_data = study_df, reference_data = reference_df,
group_var = "condition", score_var = "outcome",
items = c("item1", "item2", "item3", "item4"),
group_levels = c(reference = "control", focal = "treatment"))
# Denominator diagnostics (Table F.1 format)
diag <- denominator_diagnostics(result)
# Six-denominator sensitivity profile
sens <- denominator_sensitivity(result)
# Multi-site / meta-analysis
multisite <- multisite_ltg_smd(
data = ml_data,
site_var = "site",
group_var = "condition",
items = paste0("item", 1:25),
reference_strategy = "pooled_across_sites",
min_n_per_group = 50
)The latent true-score and target-population anchored geometric standardized mean difference (LTG-SMD) is
\[\delta_{\mathrm{LTG}} = \frac{\mu_{T1} - \mu_{T0}}{(\sigma^2_{T1,R}\,\sigma^2_{T0,R})^{1/4}},\]
a two-group SMD whose denominator is the geometric mean of group-specific true-score standard deviations in an explicitly chosen target reference population \(R\). The framework treats the denominator as an estimand choice, in the spirit of Lundberg, Johnson, and Stewart (2021), rather than as a technical detail.
| Function | Purpose |
|---|---|
compute_ltg_smd() |
Plug-in estimator + five SMD comparators (Hedges’s g, Welch, observed geometric, external observed, LTG-SMD) |
ltg_smd_ci() |
Analytic delta-method + BC/BCa bootstrap confidence intervals |
denominator_diagnostics() |
Table F.1 diagnostics for reporting (Section 8.3) |
denominator_sensitivity() |
Profile across six symmetric denominators (Section 7 Simulation D) |
sensitivity_reference() |
Sensitivity to alternative target reference distributions |
multisite_ltg_smd() |
Multi-site wrapper with cross-site reference and meta-analytic pooling |
export_supplementary() |
Bundle diagnostics, CIs, and sensitivity for a supplementary appendix |
coef_alpha() |
Cronbach’s coefficient alpha (internal default reliability estimator) |
The package vignette
vignette("getting-started", package = "ltgsmd") works
through the three Section 8 examples from the companion paper.
The three empirical illustrations in Section 8 of the paper use:
R scripts that reproduce each example are archived in the supplementary repository (see the companion paper’s Data and Code Availability statement). The scripts read the public data and call this package’s functions in the order documented in the vignette.
Please cite the companion paper:
Nakamura, D. (in press). The Denominator Chooses the Estimand: A Target-Population True-Score Framework for Standardized Mean Differences. Psychological Methods.
A CITATION file will be updated with full volume/page
details once assigned.
MIT. See the LICENSE file.
multisite_ltg_smd())This is version 0.2.2. The companion paper has been accepted for publication in Psychological Methods, and the API is stable.