Adds tdcs_paper_lme replicating the paper's exact formula/variable names (paper_*
switch cases): mergeA2 reproduces the interaction F(1)=7.09 vs 7.12, p=0.009 vs
0.008, with a coverage-confound warning. Rewrites tdcs_power_sim to fit that same
formula (interaction term canonicalized to day:stim). Adds a replication test.
Suite 37/37.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds make_figure (2-panel learning curves + per-phase slopes, colorblind-safe
Box-B2/Box-A2 palette) -> results/figure_learning.png; analysis/writeup.md
(figure + methods + results summary); and tdcs_export_all/run_export exporting
every data variation into an organized export/ tree (curated_all, scenarios/9,
anchors/3 with tDCS factor, phases/9 + MANIFEST). Adds tExport tests. Suite 36/36.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds tdcs_phase_lme (same LME refit within learning phases 0-5/6-10/6-13, with
per-phase slopes + interaction 95% CI) as phase_* switch cases, and
tdcs_power_sim (Monte-Carlo power for the early-phase interaction, scored by the
cluster-honest per-animal test and the LME test) + run_power. Findings: the
Box-B2 faster-acquisition signal is in the early phase; at n=3-5/group honest
power is 0.3-0.7 even at the observed effect (need ~8/group if effect is as
observed, ~20/group if half). Adds tPhasePower tests. Suite 35/35.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Box-A2 has no data past ~day 13, so the full-range days x tDCS interaction
extrapolates A2's slope and is confounded (spuriously significant, wrong sign).
Add the _d0_13 window (equal anchor coverage) and a report warning when the two
groups' day coverage is unequal. In the fair 0-13 window the interaction is n.s.
with a POSITIVE slope diff and the Day-1 tDCS effect is n.s. -- matching the
paper's 'equal on Day 1' direction (underpowered at n=8). Suite 32/32.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
The model already evaluates the tDCS main effect at day 0, which is the paper's
'Day 1'; relabel the report/docstring to state the equivalence explicitly so the
'equal on Day 1' comparison is unambiguous. No model change (re-indexing day
would move the main-effect evaluation point off Day 1). Suite 31/31.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Adds 6 switch cases replicating the paper's linear mixed model
success ~ day * tDCS + (1|subject) on the Box-B2(tDCS=1) vs Box-A2(tDCS=0)
arm, reporting the days x tDCS interaction, days, and tDCS effects as t/F/p.
mergeA2_full reproduces the paper's interaction (F=7.09 vs 7.12, p=0.009 vs
0.008); interpretation is sign-aware and honestly notes our data diverges from
the paper (Box-B2 ahead on day 1, gap narrows -- not 'equal on day 1,
accumulates'). Adds tLme tests. Suite 31/31.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
fitglme coefTest only offers observation-level DF (df2~=178), which overstates
the group x day interaction significance for this few-subject design (p~=0 in
every scenario). Replace the reported learning-rate result with a cluster-honest
per-animal slope test (per-subject OLS slope of success vs day, Welch + MWU,
B2 vs A2) -- now correctly parallel in all 6 scenarios, matching the Python GEE.
The GLMM F-test is kept as a flagged reference only. Adds a regression test.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Poisson GEE / binomial rate / learning-rate models, subject random-intercept
sensitivity, unknown-group classification and merge scenarios, plus METHODS.md
and learning-curve plots. Data reconstructed from the experiment DB and
verified against the exported matrix.
Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
Experiments DB Dev