Simulation Code for Conditional-Mean Inference in Random-Lot Stability Models

This Mendeley Data repository contains the worked-example dataset and the analysis/simulation code supporting the paper “Degrees-of-Freedom Approximations for Conditional-Mean Inference in Random-Lot Stability Analysis” (Karl, Rushing, Burdick, Hofer).

Worked example (cross-software illustration): The repository includes the worked-example dataset (stability_example_data.csv and stability_example_data.jmp) used to illustrate how denominator degrees-of-freedom (DDF) choices affect lot-specific conditional-mean confidence limits in random-lot stability mixed models. The JMP table includes three attached table scripts (“Default DDF”, “Contain DDF”, and a “Base JMP (Pro not required)” script) that reproduce the conditional profilers when opened and run. These JMP scripts fit the random-intercept model (Model 2); users can optionally add Lot*Months as a random effect in the JMP Fit Model dialog to explore the random intercept-and-slope model (Model 1). A SAS single-run script (stability_single_run.sas) reproduces the corresponding SAS PROC MIXED fits and conditional-mean limits. A convenience output table (Fig_2_Lot_G_bands.csv) provides the exported conditional-mean bands for the Lot G profiler illustration.

Simulation study (operating characteristics and diagnostics): The repository also contains the SAS simulation programs used to generate the paper’s main simulation results (pass/fail decision curves at a proposed expiry, coverage diagnostics, and supporting summaries). The simulation code is organized into folders for the main decision simulations, coverage simulations, margin/df export runs, and a sensitivity “spotcheck” run (details are noted in the included simulation-code readme).

All scripts are intended to be runnable as provided, with key model settings matching the paper (PROC MIXED, bounded variance components, and DDF method options such as CONTAIN vs SAT).