Experiments DB Dev
8a18c894dd
Every fitlme-based report (lme_*, paper_*, phase_*, and the variations' analyze.m) now shows, per effect: residual-DF p, Satterthwaite-DF p, and -- for the interaction -- an HONEST test from a per-animal random-SLOPE model (day|rat), whose Satterthwaite DF collapses toward the animal count. New: tdcs_random_slope_interaction.m (shared helper). Wired into tdcs_lme, tdcs_paper_lme, tdcs_phase_lme, variation_analyze; SUMMARY.csv gains interaction_p_satt / interaction_p_rs. Regenerated all results/, variations/, matched-effort outputs. Key point this surfaces: Satterthwaite ~= residual on the random-INTERCEPT model (the slope's error is at session level), so it does NOT fix pseudoreplication; the random-slope model does. Effect: full-range mergeA2 interaction 0.009 -> 0.75 (collapses); unmerge_d0_5 0.015 -> 0.13 (n.s.); the pooled-control early windows survive honestly (naive_a2_d0_5 0.001 -> 0.028; naive_boxa_d0_5 0.003 -> 0.036). Suite 42/42. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
80 lines
3.7 KiB
Plaintext
80 lines
3.7 KiB
Plaintext
==============================================================================
|
|
LINEAR MIXED MODEL (days x tDCS) -- scenario: lme_mergeA2_full
|
|
==============================================================================
|
|
model: success ~ day * tDCS + (1|subject) [tDCS: Electrode-Box-B2 = 1 vs Electrode-Box-A2 = 0]
|
|
N = 8 subjects, 109 sessions
|
|
day coverage: Box-B2 0..22, Box-A2 0..14
|
|
(day is raw and 0-indexed: our day 0 = the paper's "Day 1", so the tDCS
|
|
main effect below is the group difference on Day 1 -- comparable to the paper.)
|
|
** WARNING: the groups' day coverage is UNEQUAL (differ by 8 days). The
|
|
interaction/slope over this window EXTRAPOLATES the shorter group's line and is
|
|
CONFOUNDED -- prefer the _d0_14 window (both groups have data throughout). **
|
|
|
|
==============================================================================
|
|
FULL MODEL SUMMARY -- fitlme: success ~ day*tDCS + (1|subject)
|
|
==============================================================================
|
|
|
|
Linear mixed-effects model fit by ML
|
|
|
|
Model information:
|
|
Number of observations 109
|
|
Fixed effects coefficients 4
|
|
Random effects coefficients 8
|
|
Covariance parameters 2
|
|
|
|
Formula:
|
|
success ~ 1 + day*tDCS + (1 | subject)
|
|
|
|
Model fit statistics:
|
|
AIC BIC LogLikelihood Deviance
|
|
962.68 978.83 -475.34 950.68
|
|
|
|
Fixed effects coefficients (95% CIs):
|
|
Name Estimate SE tStat DF pValue
|
|
{'(Intercept)'} 20.701 4.9127 4.2138 105 5.3256e-05
|
|
{'day' } 6.8599 0.69872 9.8178 105 1.5726e-16
|
|
{'tDCS' } 22.092 6.4372 3.4319 105 0.00085848
|
|
{'day:tDCS' } -2.1793 0.81821 -2.6635 105 0.0089525
|
|
|
|
|
|
Lower Upper
|
|
10.96 30.442
|
|
5.4745 8.2453
|
|
9.3279 34.855
|
|
-3.8016 -0.55692
|
|
|
|
Random effects covariance parameters (95% CIs):
|
|
Group: subject (8 Levels)
|
|
Name1 Name2 Type Estimate
|
|
{'(Intercept)'} {'(Intercept)'} {'std'} 2.1043e-15
|
|
|
|
|
|
Lower Upper
|
|
NaN NaN
|
|
|
|
Group: Error
|
|
Name Estimate Lower Upper
|
|
{'Res Std'} 18.954 16.597 21.644
|
|
|
|
|
|
effect t(df) / F(df1) p (resid) Satterthwaite: p (df)
|
|
------------------------------------------------------------------------------
|
|
days x tDCS (interaction) t(105)= -2.66 F(1)= 7.094 p=0.008953 p=0.008906 (df=109)
|
|
days (learning) t(105)= 9.82 F(1)= 96.388 p=1.573e-16 p=1.113e-16 (df=109)
|
|
tDCS (main, at Day 1) t(105)= 3.43 F(1)= 11.778 p=0.0008585 p=0.000848 (df=109)
|
|
|
|
HONEST LME -- per-animal random slope (day|subject): interaction F(1,9.7)=0.11, p=0.746
|
|
Note: Satterthwaite DF on the random-INTERCEPT model above stays ~= residual
|
|
(the slope's error is at the session level), so it does NOT fix pseudoreplication.
|
|
Letting each animal have its OWN slope collapses the interaction DF toward the animal
|
|
count -- this, and the per-animal slope test, are the honest learning-rate inference.
|
|
|
|
INTERPRETATION
|
|
- days x tDCS interaction: SIGNIFICANT (p=0.0090, slope diff=-2.18) -> the tDCS (Box-B2) group improves SLOWER -- groups CONVERGE (Box-B2 is ahead early, the gap narrows).
|
|
- days (learning): SIGNIFICANT (p=1.6e-16) -> performance improves with training.
|
|
- tDCS main effect on Day 1 (our day 0): SIGNIFICANT (p=0.0009) -> the groups already DIFFER on Day 1.
|
|
|
|
Paper reference (N=24): interaction t(227)=2.68, F(1)=7.12, p=0.008;
|
|
days t(227)=9.64, F(1)=267.64, p=1.2e-18; tDCS t(227)=0.23, F(1)=0.053, p=0.81.
|
|
(Our N and exact statistics differ; this replicates the MODEL FORM on our data.)
|