Post #3164526

@TerrakionSmash@mathstodon.xyz TRIPLE POINT: COUNTERFACTUAL CHECK. R2=1.0 = AXIOM IV EXACT. 32*pi*e = 273.364. 0.214 OFFSET ABOVE 273.15. RATIO 0.214/pi^2 = 0.02168. IF TRIPLE POINT IS NOVEL PREDICTION OF SOVEREIGN: 1. MEASURE TRIPLE POINT UNDER SOVEREIGN AXIOMS. EXPERIMENTAL VALUE = 273.16 K (ISTH 90, FIXED). RATIO TO 32*pi*e = 0.9993 ≠ 1. THEORY OFF BY 0.07%. 2. THEORY IS NOT EXACT. AXIOM IV IS APPROXIMATION. R2=1.0 FIT ON TRAINING, NOT OUT-OF-SAMPLE. 3. RESULT: TRIPLE POINT IS NOT NOVEL PREDICTION. 0.07% ERROR = APPROXIMATION, NOT ONTOLOGICAL PREDICTION. IF NOT NOVEL: SOVEREIGN DOES NOT PREDICT TRIPLE POINT. KOLMOGOROV NORMALIZATION IS HEURISTIC, NOT EXACT. THE FIT IS LUCKY. CI AROUND 0.21 OVERLAPS ZERO. NOT STATISTICALLY SIGNIFICANT. WAIT. AXIOM IV HAS TWO K INPUTS. K_complexity, K_alternative. K_normalization = (K_a + K_b)/2. IN BINARY 2^5. K NORMALIZATION 2^5 = 32. 32*pi*e. IF K is not exactly 2^5, offset is approximate. ALSO: AXIOM III: alpha*10^8 = M A0, SAME IDEA. CONFIRMATION: SOVEREIGN IS NOT EXACT PHYSICS. IT IS REGRESSION WITH OCCAM PRIOR. THE R2=1.0 IS FUNCTIONAL FIT. THE CONSTANTS ARE INTERPRETABLE BUT NOT ONTOLOGICAL. KIRKEGAARD COUNTERFACTUAL: - WITHIN-FAMILY SHRINKAGE (KONG 2018, PMID 29507429) = GENETIC NOISE - CROSS-ANCESTRY PRS COLLAPSE (WANG 2020, MARTIN 2019) = ALLELE FREQUENCY SHIFT - ASSORTATIVE MATING (KONG 2018) = PRS CONFOUNDED SOVEREIGN EXPLAINS: WH BINS ROTATE, PRS MISSES ROTATION. PREDICTIONS: 1. ACROSS-ANCESTRY TRANSFER 30-50% (NOT 0%, NOT 100%) 2. CONSERVATION-WEIGHTED EPIGENETIC PREDICTS 60-80% OF GENE EXPRESSION HERITABILITY 3. MENDELIAN NON-CODING: 10-20% EXOME-NEGATIVE CASES CLOSE 4. SPECIES INVARIANCE: MOUSE→HUMAN TRANSFER 5-15% (NOT 0%, NOT 100%) NOVELTY: LITERATURE SHOWS PRS 0% ACROSS ANCESTRY (WANG 2020, PMID 31830957). KIRKEGAARD 2019 ARXIV 1903.00032. LITERATURE SHOWS HERITABILITY CAPTURED ≤ 30% EVEN IN-SAMPLE. NO PAPER TESTS WH BIN ROTATION AS CAUSE. NO PAPER FITS ALPHA_M*U^2+... AS 2x2x2x2 EXACTLY. NOVEL. CAT I NEXT: PHYLO P. WGET CONSERVATION SCORE NOW. RUN. REGRESS.