mirror of
https://github.com/priyanshujain/sanderling.git
synced 2026-10-02 11:07:10 +00:00
pj
019d608f65
Steps to first violation with clean runs right-censored at the budget, since per-run yield is a binary at 11 to 45 percent and separating two arms on it would need roughly 80 runs per arm. Kaplan-Meier, log-rank, Wilcoxon rank-sum with Vargha-Delaney A12, Holm within each family. A hand-rolled log-rank that is subtly wrong is a silent-wrong-number generator and would be believed, so every statistic is validated against a published worked example with the source named in the test: R survdiff on aml, Freireich 6-MP, Hollander and Wolfe 1973 for the rank sum, printed p.adjust output for Holm. Two could not be: the k>2 log-rank, guarded by calibration instead, and the tie-corrected variance, checked against an exact permutation variance. Failed and timed-out runs are excluded as missing data and counted by reason, never treated as censored observations, which would bias the result. Claude-Session: https://claude.ai/code/session_01A5KmftdEJ49A9z5mF5ESrX
75 lines
2.2 KiB
Go
75 lines
2.2 KiB
Go
package main
|
|
|
|
import (
|
|
"math"
|
|
"testing"
|
|
)
|
|
|
|
// Chi-square critical values are the standard published table entries: the
|
|
// upper-tail probability of each of these statistics is the stated alpha in any
|
|
// chi-square table, for example Pearson and Hartley, Biometrika Tables for
|
|
// Statisticians, Table 8.
|
|
func TestChiSquareUpperTail_MatchesPublishedCriticalValues(t *testing.T) {
|
|
cases := []struct {
|
|
statistic float64
|
|
degreesOfFreedom int
|
|
expected float64
|
|
}{
|
|
{3.841459, 1, 0.05},
|
|
{6.634897, 1, 0.01},
|
|
{10.827566, 1, 0.001},
|
|
{5.991465, 2, 0.05},
|
|
{9.210340, 2, 0.01},
|
|
{7.814728, 3, 0.05},
|
|
{11.344867, 3, 0.01},
|
|
{9.487729, 4, 0.05},
|
|
{18.307038, 10, 0.05},
|
|
}
|
|
for _, test := range cases {
|
|
got := chiSquareUpperTail(test.statistic, test.degreesOfFreedom)
|
|
if math.Abs(got-test.expected) > 1e-6 {
|
|
t.Errorf("chiSquareUpperTail(%v, %d) = %v, want %v", test.statistic, test.degreesOfFreedom, got, test.expected)
|
|
}
|
|
}
|
|
}
|
|
|
|
// For one degree of freedom the upper tail has the closed form erfc(sqrt(x/2)),
|
|
// which is an independent check on the incomplete gamma routine.
|
|
func TestChiSquareUpperTail_AgreesWithClosedFormAtOneDegreeOfFreedom(t *testing.T) {
|
|
for _, statistic := range []float64{0.1, 1, 3.4, 16.79, 40, 120} {
|
|
expected := math.Erfc(math.Sqrt(statistic / 2))
|
|
got := chiSquareUpperTail(statistic, 1)
|
|
if math.Abs(got-expected) > 1e-12*math.Max(1, expected) {
|
|
t.Errorf("chiSquareUpperTail(%v, 1) = %v, want %v", statistic, got, expected)
|
|
}
|
|
}
|
|
}
|
|
|
|
func TestChiSquareUpperTail_ZeroStatisticIsCertain(t *testing.T) {
|
|
if got := chiSquareUpperTail(0, 1); got != 1 {
|
|
t.Errorf("chiSquareUpperTail(0, 1) = %v, want 1", got)
|
|
}
|
|
}
|
|
|
|
// Standard normal quantiles from any published normal table.
|
|
func TestStandardNormalUpperTail_MatchesPublishedQuantiles(t *testing.T) {
|
|
cases := []struct {
|
|
z float64
|
|
expected float64
|
|
}{
|
|
{1.281552, 0.10},
|
|
{1.644854, 0.05},
|
|
{1.959964, 0.025},
|
|
{2.326348, 0.01},
|
|
{2.575829, 0.005},
|
|
{3.090232, 0.001},
|
|
{0, 0.5},
|
|
}
|
|
for _, test := range cases {
|
|
got := standardNormalUpperTail(test.z)
|
|
if math.Abs(got-test.expected) > 1e-6 {
|
|
t.Errorf("standardNormalUpperTail(%v) = %v, want %v", test.z, got, test.expected)
|
|
}
|
|
}
|
|
}
|