Conversation
Add a k=3 fraction-of-design-space study contrasting the design families from Igor Miranda Santana's central composite design post, in the style of the book's omnibus comparison across design families. - ccd_variants_designs.py: shared builders (circumscribed, face-centred and inscribed CCDs via direct construction; Box-Behnken via process_improve; a canonical simplex-difference Doehlert generator) and an evaluate() wrapper over process_improve's FDS metric on the full second-order model. - fds-compare-design-families.py: overlaid FDS (scaled prediction variance) for the five families over the cube, linear and log panels. - fds-ccd-varying-alpha.py: FDS of one CCD as the axial distance alpha is swept from face-centred (1) through rotatable (1.682) and beyond. - ccd-variants-geometry.py: k=2 coded-space geometry mirroring the post. - check_ccd_variants.py: construction assertions plus the k=3 comparison table and the k=5 run counts (43/27/41/31) quoted in the post. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PEPUURkGJ1QJK5F2KZZeH8
The uniform-shell designs (Doehlert, inscribed CCD) only look bad on the cube because their reach stops at the unit sphere; over the ball they are built for they are competitive. Add ball_fds_curve() to integrate the scaled prediction variance over a chosen radius (the library's spherical region uses the wider radius-sqrt(k) circumscribing sphere), and a side-by-side figure showing the ranking flip between the two regions. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01PEPUURkGJ1QJK5F2KZZeH8
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
2 participants
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.
What this adds
A fraction-of-design-space (FDS) study contrasting the design families from Igor Miranda Santana's central composite design post (linked below), built in the style of the book's omnibus comparison across design families. All new files live in
doe/.The comparison runs at k = 3 factors - the smallest number at which every family in the post exists (Box-Behnken needs k >= 3) and at which all of them are full rank on the complete second-order model (intercept + 3 linear + 3 two-factor interactions + 3 pure quadratics = 10 terms).
New files
ccd_variants_designs.py- shared builders and anevaluate()wrapper, mirroringomnibus_designs.py. The central composite variants (circumscribed/rotatable, face-centred, inscribed) are built by direct construction, because the library's full-cube CCD only honours the named alphas; Box-Behnken comes fromprocess_improve; Doehlert uses a canonical simplex-difference generator (all shell points on the unit sphere). Prediction variance and the dense FDS curve come fromprocess_improve'sevaluate_designover the cube region with the corner vertices included.fds-compare-design-families.py->fds-compare-design-families.png. Overlaid FDS (scaled prediction variance) for the five families, linear panel for the central bulk and log panel out to the corners. The inscribed CCD and Doehlert never leave the unit sphere, so they extrapolate badly at the cube corners; the circumscribed and face-centred CCDs and Box-Behnken stay well behaved.fds-ccd-varying-alpha.py->fds-ccd-varying-alpha.png. FDS of one CCD as the axial distance alpha is swept from face-centred (1) through rotatable (1.682) and out to 2, showing the interior-variance-versus-reach trade-off.ccd-variants-geometry.py->ccd-variants-geometry.png. A k = 2 coded-space picture of the three CCD variants and Doehlert, mirroring the diagram in the post.check_ccd_variants.py- construction assertions, the k = 3 comparison table the figures visualise, and the k = 5 run counts (43 / 27 / 41 / 31) quoted in the post, reproduced and asserted.Reproducing
From
doe/:python3 check_ccd_variants.pyvalidates the constructions and prints both tables; the three*.pyfigure scripts each regenerate their PNG. Requiresprocess-improve[expt](used as a library; no changes to that package).Source post: https://www.linkedin.com/posts/igor-miranda-santana-64090b53_doe-activity-7482811159092580352-IrfX
🤖 Generated with Claude Code
https://claude.ai/code/session_01PEPUURkGJ1QJK5F2KZZeH8
Generated by Claude Code