Repository navigation
Make Satterthwaite the default inference of fit_solomon_ml() (#115) - #124
Merged
Merged
Conversation
fit_solomon_ml() now defaults to inference = "satterthwaite". Point estimates are unchanged; standard errors, intervals, and p-values change for code that relied on the old default, which can reproduce earlier results with inference = "wald". "wald" remains, documented as van Engelenburg's (1999) large-sample inference. The package's simulation study (#10, #22) found Wald intervals liberal in small samples (mean coverage .920 at 10 per cell) and Satterthwaite inference calibrated at every cell size studied. - The classed small-sample warning (solomonR_small_sample_warning) is removed. It recommended what is now the default and was silent whenever "wald" was chosen explicitly, now the only way to get Wald inference. The printed note for Wald fits with fewer than 40 per cell stays and now states the threshold. - compare_solomon_methods() lists "Maximum likelihood (Satterthwaite)" first and "Maximum likelihood (Wald)"; the rows were "Maximum likelihood (small-sample)" and "Maximum likelihood". - equivalence_solomon() labelled every ML fit "large-sample normal reference", even Satterthwaite fits tested with t; it now uses the fit's inference, through a shared .ml_inference_label() also used by figure captions. - Printed output, captions, and report_solomon() say "Satterthwaite inference" instead of "the small-sample option"; report_solomon() states which contrasts use residual and which Welch-Satterthwaite df, and calls Wald tests large-sample. Tests: existing expectations changed because the behavior was changed on purpose, not because they were wrong before: - test-ml-inference.R: "small cells warn only when inference is not chosen" and "cells at the threshold do not warn" are replaced by a test that no inference choice warns; the print test now expects "Inference: Satterthwaite" (was "Inference: small-sample"). - test-plot-effects.R and test-plot-sensitization.R expect the caption "maximum likelihood; Satterthwaite inference" (was "small-sample option"). - test-compare.R selects rows by the new ML labels. New tests cover the default, the absence of the warning, the printed note at and below the threshold, the report text and references for both options, the equivalence label and df for Satterthwaite fits, and the order and values of the two ML rows in compare_solomon_methods(). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…les (#115) - Getting-started guide: the default ML fit is printed, and Wald inference is shown as inference = "wald" with its small-cell note; the table of analyses no longer says to choose "satterthwaite" for small groups. - Method guide, GLM vignette, README (re-knitted), and the several- treatments article describe Satterthwaite inference as the default and Wald inference as van Engelenburg's (1999) large-sample option, with Satterthwaite (1946) and Welch (1947) cited and listed where they are named. - ml-validation article: notes that the run predates the change and named every fit's inference, so its results stand; renames the warning threshold as the small-cell threshold for Wald inference; shows the methods as "ML, Satterthwaite (default)" and "ML, Wald (van Engelenburg, 1999)" while performance.csv keeps the run's labels. - validation-evidence: build-benchmarks.R applies the same labels; benchmarks.csv and studies.csv are regenerated (only the ML method labels change; an unmodified rebuild reproduced the committed files exactly). - References page: the van Engelenburg (1999) note records the default. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Review follow-up for #115. - The Satterthwaite label of .ml_inference_label() and the print header of fit_solomon_ml() now cite (Satterthwaite, 1946; Welch, 1947), as the Wald label cites van Engelenburg (1999). equivalence_solomon() wraps the longer label to the console width. - With the citation, a caption line of plot_solomon_effects() and plot_sensitization() reached 140 characters and was cut off at 7 inches. The inference now has a line of its own, after the confidence level and reference distribution; tests check that every caption line fits 92 characters, the count for a 7-inch figure at 0.55 em per character. - The printout labels the residual SDs as ML estimates (SSE / n) and, for Satterthwaite fits, also gives the SDs from the unbiased variances (SSE / df) that its SEs use, stored in the new sigma_unbiased field. Tests tie each printed SD to the printed SE it produces. - ?compare_solomon_methods lists Satterthwaite (1946) and Welch (1947), and ?plot_solomon_effects lists them and van Engelenburg (1999), since their output cites these works; a test checks the help pages. Changed expectations: test-ml-inference.R, test-plot-effects.R, and test-plot-sensitization.R expected the Satterthwaite label without its sources, which the label now carries on purpose. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This was referenced Oct 4, 2026
Merged
NEWS: this branch's sections go under the development heading, above the 0.8.1 section. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
R/solomon_plot_effects.R: the ML entry keeps this branch's cited inference label and gains master's bounds scale; the references are the union, in order; the caption keeps the inference on its own line and names the thin bars when bounds are drawn. The TOST outcome now goes on a line of its own, as it does for designs with several treatments, so every caption line fits a 7-inch figure. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
NEWS: this branch's section goes first under the development heading. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
NEWS: this branch's section goes first under the development heading. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
JUhalt
added a commit
that referenced
this pull request
Oct 4, 2026
NEWS: this branch's sections go first under the development heading. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This was referenced Oct 8, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Fixes #115, as the maintainer decided on October 3, 2026.
fit_solomon_ml(inference = "satterthwaite")is now the default (Satterthwaite, 1946; Welch, 1947)."wald"remains, labelled as van Engelenburg's (1999) large-sample inference. The package's own study found the Wald default liberal (coverage .920 at 10 per cell), while the Satterthwaite option was calibrated ([v0.3] Broaden fit_solomon_ml simulation validation #10, [Proposal] Small-sample inference for fit_solomon_ml() #22).compare_solomon_methods(),plot_sensitization(), thereport_solomon()text), the vignettes, the README, the validation article, and the roadmap. NEWS calls this a change of default that can alter results.sigma_unbiased.plot_solomon_effects()andplot_sensitization()put the inference on a line of its own, so the longer cited label fits a 7-inch figure. Tests check every line fits.Checks
tools/check-references.Rand the spelling check are clean.Review. An independent reviewer checked the branch against the issues' acceptance criteria and re-ran its statistics. Every confirmed finding was then fixed, with tests.
References
Satterthwaite, F. E. (1946). An approximate distribution of estimates of variance components. Biometrics Bulletin, 2(6), 110–114. https://doi.org/10.2307/3002019
van Engelenburg, G. (1999). Statistical analysis for the Solomon four-group design (Research Report 99-06). University of Twente. ERIC. https://eric.ed.gov/?id=ED435692
Welch, B. L. (1947). The generalization of "Student's" problem when several different population variances are involved. Biometrika, 34(1–2), 28–35. https://doi.org/10.1093/biomet/34.1-2.28
🤖 Generated with Claude Code