Upload results/ralph23_variance_decomposition_summary.txt with huggingface_hub
Browse files
results/ralph23_variance_decomposition_summary.txt
ADDED
|
@@ -0,0 +1,166 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
================================================================================
|
| 2 |
+
Ralph23 Task 5: Partial R-squared Variance Decomposition
|
| 3 |
+
Date: 2026-02-12
|
| 4 |
+
Script: ralph23_t05_variance_decomposition_20260212_000354.py
|
| 5 |
+
Runtime: 1.0 seconds
|
| 6 |
+
================================================================================
|
| 7 |
+
|
| 8 |
+
Method: OLS linear regression with per-fold standardization
|
| 9 |
+
CV scheme: 9-fold spatial CV (from ralph23_spatial_cv_folds.npz)
|
| 10 |
+
Bio-valid samples: 1151 / 1810
|
| 11 |
+
|
| 12 |
+
|
| 13 |
+
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 14 |
+
PFAM Dimensionality: pfam20
|
| 15 |
+
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 16 |
+
|
| 17 |
+
Chl-a:
|
| 18 |
+
RΒ²_env (mean Β± sd): -48.5449 Β± 134.3695
|
| 19 |
+
RΒ²_pfam (mean): -11.5829
|
| 20 |
+
RΒ²_env+pfam (mean Β± sd): -48.5299 Β± 134.7549
|
| 21 |
+
Unique PFAM (mean Β± sd): +0.0150 Β± 0.7285
|
| 22 |
+
Unique Env (mean): -36.9470
|
| 23 |
+
Shared (mean): -11.5979
|
| 24 |
+
ΞRΒ² (both - env): +0.0150
|
| 25 |
+
Paired t-test: t=0.062, p=0.9522
|
| 26 |
+
Wilcoxon signed-rank: p=0.9102
|
| 27 |
+
Cohen's d: 0.0206
|
| 28 |
+
95% CI for ΞRΒ²: [-0.5450, +0.5750]
|
| 29 |
+
Folds +/-/0: 4+/5-/0=
|
| 30 |
+
|
| 31 |
+
POC:
|
| 32 |
+
RΒ²_env (mean Β± sd): -6.1351 Β± 10.7978
|
| 33 |
+
RΒ²_pfam (mean): -5.2632
|
| 34 |
+
RΒ²_env+pfam (mean Β± sd): -6.1092 Β± 10.3026
|
| 35 |
+
Unique PFAM (mean Β± sd): +0.0259 Β± 0.5680
|
| 36 |
+
Unique Env (mean): -0.8460
|
| 37 |
+
Shared (mean): -5.2891
|
| 38 |
+
ΞRΒ² (both - env): +0.0259
|
| 39 |
+
Paired t-test: t=0.137, p=0.8947
|
| 40 |
+
Wilcoxon signed-rank: p=0.1289
|
| 41 |
+
Cohen's d: 0.0455
|
| 42 |
+
95% CI for ΞRΒ²: [-0.4108, +0.4625]
|
| 43 |
+
Folds +/-/0: 1+/8-/0=
|
| 44 |
+
|
| 45 |
+
NFLH:
|
| 46 |
+
RΒ²_env (mean Β± sd): 0.9556 Β± 0.0426
|
| 47 |
+
RΒ²_pfam (mean): -0.2838
|
| 48 |
+
RΒ²_env+pfam (mean Β± sd): 0.9550 Β± 0.0426
|
| 49 |
+
Unique PFAM (mean Β± sd): -0.0006 Β± 0.0063
|
| 50 |
+
Unique Env (mean): +1.2388
|
| 51 |
+
Shared (mean): -0.2832
|
| 52 |
+
ΞRΒ² (both - env): -0.0006
|
| 53 |
+
Paired t-test: t=-0.296, p=0.7750
|
| 54 |
+
Wilcoxon signed-rank: p=0.4258
|
| 55 |
+
Cohen's d: -0.0986
|
| 56 |
+
95% CI for ΞRΒ²: [-0.0055, +0.0043]
|
| 57 |
+
Folds +/-/0: 3+/6-/0=
|
| 58 |
+
|
| 59 |
+
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 60 |
+
PFAM Dimensionality: pfam32
|
| 61 |
+
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 62 |
+
|
| 63 |
+
Chl-a:
|
| 64 |
+
RΒ²_env (mean Β± sd): -48.5449 Β± 134.3695
|
| 65 |
+
RΒ²_pfam (mean): -16.0561
|
| 66 |
+
RΒ²_env+pfam (mean Β± sd): -38.7935 Β± 101.8234
|
| 67 |
+
Unique PFAM (mean Β± sd): +9.7514 Β± 32.6850
|
| 68 |
+
Unique Env (mean): -22.7374
|
| 69 |
+
Shared (mean): -25.8075
|
| 70 |
+
ΞRΒ² (both - env): +9.7514
|
| 71 |
+
Paired t-test: t=0.895, p=0.3969
|
| 72 |
+
Wilcoxon signed-rank: p=0.3008
|
| 73 |
+
Cohen's d: 0.2983
|
| 74 |
+
95% CI for ΞRΒ²: [-15.3725, +34.8753]
|
| 75 |
+
Folds +/-/0: 2+/7-/0=
|
| 76 |
+
|
| 77 |
+
POC:
|
| 78 |
+
RΒ²_env (mean Β± sd): -6.1351 Β± 10.7978
|
| 79 |
+
RΒ²_pfam (mean): -6.6967
|
| 80 |
+
RΒ²_env+pfam (mean Β± sd): -9.1903 Β± 16.1200
|
| 81 |
+
Unique PFAM (mean Β± sd): -3.0553 Β± 5.4701
|
| 82 |
+
Unique Env (mean): -2.4937
|
| 83 |
+
Shared (mean): -3.6414
|
| 84 |
+
ΞRΒ² (both - env): -3.0553
|
| 85 |
+
Paired t-test: t=-1.676, p=0.1323
|
| 86 |
+
Wilcoxon signed-rank: p=0.0039
|
| 87 |
+
Cohen's d: -0.5585
|
| 88 |
+
95% CI for ΞRΒ²: [-7.2600, +1.1495]
|
| 89 |
+
Folds +/-/0: 0+/9-/0=
|
| 90 |
+
|
| 91 |
+
NFLH:
|
| 92 |
+
RΒ²_env (mean Β± sd): 0.9556 Β± 0.0426
|
| 93 |
+
RΒ²_pfam (mean): -0.3600
|
| 94 |
+
RΒ²_env+pfam (mean Β± sd): 0.9573 Β± 0.0371
|
| 95 |
+
Unique PFAM (mean Β± sd): +0.0017 Β± 0.0082
|
| 96 |
+
Unique Env (mean): +1.3173
|
| 97 |
+
Shared (mean): -0.3617
|
| 98 |
+
ΞRΒ² (both - env): +0.0017
|
| 99 |
+
Paired t-test: t=0.633, p=0.5445
|
| 100 |
+
Wilcoxon signed-rank: p=0.9102
|
| 101 |
+
Cohen's d: 0.2109
|
| 102 |
+
95% CI for ΞRΒ²: [-0.0046, +0.0080]
|
| 103 |
+
Folds +/-/0: 4+/5-/0=
|
| 104 |
+
|
| 105 |
+
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 106 |
+
PFAM Dimensionality: pfam64
|
| 107 |
+
ββββββββββββββββββββββββββββββββββββββββββββββββββββββββββββ
|
| 108 |
+
|
| 109 |
+
Chl-a:
|
| 110 |
+
RΒ²_env (mean Β± sd): -48.5449 Β± 134.3695
|
| 111 |
+
RΒ²_pfam (mean): -21.2594
|
| 112 |
+
RΒ²_env+pfam (mean Β± sd): -41.8854 Β± 107.5039
|
| 113 |
+
Unique PFAM (mean Β± sd): +6.6595 Β± 27.3695
|
| 114 |
+
Unique Env (mean): -20.6260
|
| 115 |
+
Shared (mean): -27.9189
|
| 116 |
+
ΞRΒ² (both - env): +6.6595
|
| 117 |
+
Paired t-test: t=0.730, p=0.4862
|
| 118 |
+
Wilcoxon signed-rank: p=0.1289
|
| 119 |
+
Cohen's d: 0.2433
|
| 120 |
+
95% CI for ΞRΒ²: [-14.3786, +27.6975]
|
| 121 |
+
Folds +/-/0: 1+/8-/0=
|
| 122 |
+
|
| 123 |
+
POC:
|
| 124 |
+
RΒ²_env (mean Β± sd): -6.1351 Β± 10.7978
|
| 125 |
+
RΒ²_pfam (mean): -9.2675
|
| 126 |
+
RΒ²_env+pfam (mean Β± sd): -9.7802 Β± 13.1164
|
| 127 |
+
Unique PFAM (mean Β± sd): -3.6451 Β± 4.2615
|
| 128 |
+
Unique Env (mean): -0.5127
|
| 129 |
+
Shared (mean): -5.6224
|
| 130 |
+
ΞRΒ² (both - env): -3.6451
|
| 131 |
+
Paired t-test: t=-2.566, p=0.0333
|
| 132 |
+
Wilcoxon signed-rank: p=0.0039
|
| 133 |
+
Cohen's d: -0.8553
|
| 134 |
+
95% CI for ΞRΒ²: [-6.9208, -0.3694]
|
| 135 |
+
Folds +/-/0: 0+/9-/0=
|
| 136 |
+
|
| 137 |
+
NFLH:
|
| 138 |
+
RΒ²_env (mean Β± sd): 0.9556 Β± 0.0426
|
| 139 |
+
RΒ²_pfam (mean): -0.6591
|
| 140 |
+
RΒ²_env+pfam (mean Β± sd): 0.9546 Β± 0.0356
|
| 141 |
+
Unique PFAM (mean Β± sd): -0.0010 Β± 0.0108
|
| 142 |
+
Unique Env (mean): +1.6137
|
| 143 |
+
Shared (mean): -0.6581
|
| 144 |
+
ΞRΒ² (both - env): -0.0010
|
| 145 |
+
Paired t-test: t=-0.266, p=0.7970
|
| 146 |
+
Wilcoxon signed-rank: p=0.5703
|
| 147 |
+
Cohen's d: -0.0886
|
| 148 |
+
95% CI for ΞRΒ²: [-0.0093, +0.0074]
|
| 149 |
+
Folds +/-/0: 3+/6-/0=
|
| 150 |
+
|
| 151 |
+
|
| 152 |
+
================================================================================
|
| 153 |
+
INTERPRETATION
|
| 154 |
+
================================================================================
|
| 155 |
+
|
| 156 |
+
POC (primary target) across PFAM dimensionalities:
|
| 157 |
+
pfam20: ΞRΒ² = +0.0259, p = 0.8947 (ns), d = 0.046, CI = [-0.4108, +0.4625]
|
| 158 |
+
pfam32: ΞRΒ² = -3.0553, p = 0.1323 (ns), d = -0.559, CI = [-7.2600, +1.1495]
|
| 159 |
+
pfam64: ΞRΒ² = -3.6451, p = 0.0333 (*), d = -0.855, CI = [-6.9208, -0.3694]
|
| 160 |
+
|
| 161 |
+
RESULT: Adding PFAM features significantly HURTS POC prediction (negative direction).
|
| 162 |
+
The additional features introduce noise that degrades OLS generalization across spatial folds.
|
| 163 |
+
|
| 164 |
+
Note: This is a model-agnostic analysis using simple OLS regression.
|
| 165 |
+
Nonlinear PFAM-env interactions would not be captured by this method.
|
| 166 |
+
The signal must be linearly detectable in the feature space for this test to detect it.
|