more output plots

[n8thangreen]

1. Covariate Balance (Love) Plots

• The Concept: Since the Matching-Adjusted Indirect Comparison (MAIC) strategy works by re-weighting individual patient data (IPD) to match the aggregate baseline characteristics of the target trial, visualizing this shift is crucial.

• The Plot: A Love plot showing the standardized mean differences (SMDs) of the covariates before and after applying the calculated weights. The text actually notes that external packages like cobalt are currently used for "diagnosing covariate balance before and after adjustment". Bringing a native version of this into outstandR would perfectly complement your strategy_maic() workflow.

2. Synthetic Cohort Density Overlays

• The Concept: For the G-computation strategies, the package generates a "synthetic cohort" of size $N$ using a Gaussian copula to match the aggregate-level data (ALD). Users can even specify custom marginal distributions like Gamma or Binomial.

• The Plot: Density plots that overlay the simulated covariate distributions of the pseudo-population against vertical lines representing the target ALD means or bounds. This would give users immediate visual confirmation that the inverse probability transforms and custom marginal parameters successfully recreated the target population.

3. Posterior Distribution Density Plots

• The Concept: The Bayesian G-computation method estimates the "full posterior distribution" of the outcome model parameters rather than just point estimates.

• The Plot: While you currently offer MCMC trace plots to check for chain convergence, adding smooth density plots (ridge plots or violin plots) for the posterior distributions of the marginal treatment effects would be a fantastic addition. It would allow analysts to make those intuitive, direct probabilistic statements (e.g., "95% probability the effect is between $a$ and $b$") at a glance.

4. Effective Sample Size (ESS) Sensitivity Curves

• The Concept: The documentation warns about a bias-variance trade-off: when users try to match higher-order moments (like variances and covariances) in MAIC, the optimizer solves for more parameters, which "commonly results in more extreme weights and a lower ESS".

• The Plot: A line chart plotting the Effective Sample Size (ESS) on the y-axis against different balance_model complexities on the x-axis (e.g., mean only, mean + variance, mean + variance + interactions). This would help users visually pinpoint exactly where their model becomes too complex for the available data.