The clinical illustration uses IBScovars from
DoseFinding 1.4-1. The source is the CRAN package
archive. The analysis retained all 369 released
observations, with dose counts 71, 78, 75, 72, and 73 at codes
0, 1, 2, 3, and 4. No observation was excluded, imputed, standardized,
or transformed for this analysis. The released outcome is a
baseline-adjusted abdominal-pain response score; larger values indicate
greater improvement according to the source documentation.
The release contains completers, so this is not an intention-to-treat reconstruction. Active physical doses and spacing are not supplied. The equal-spacing code convention is explicit and affects the integrated target.
clinical_dose_summary.tsv contains dose-level summaries
derived from the unchanged released sample.
clinical_intervals.tsv contains the saved completed
interval results. The package contains no individual-level clinical
data.
| File | Field | Meaning |
|---|---|---|
| Dose summary | dose | Ordered dose code, 0 through 4; not a physical dose |
| Dose summary | n | Number of released observations at that dose |
| Dose summary | mean | Sample mean of the released response |
| Dose summary | sd | Sample standard deviation with denominator n - 1 |
| Interval results | method / label | Recorded method identifier and display label |
| Interval results | status | Recorded numerical status |
| Interval results | estimate / qhat | Common fitted target estimate and shape estimate |
| Interval results | lower / upper / width | Reported interval endpoints and width in response units |
| Interval results | calibration | Description of the method’s calibration |
Let \(s_j\) be the dose-specific sample standard deviation and \(n_j\) its count. The pooled within-dose variance for \(m\) dose levels is computed as
\[S^2=\frac{\sum_{j=1}^m(n_j-1)s_j^2}{\sum_{j=1}^m n_j-m}.\]
This expression pools pure error under the common-variance assumption; its denominator is the pure-error degrees of freedom. It reproduces the variance used by the confidence region, up to the precision of the saved summary text. It is not estimated from deviations about a fitted Emax curve.
The source package and aggregate summaries are available from the package website and GitHub repository. Source-dataset documentation and applicable reuse conditions should accompany any future deposition.