What does Koopman Modes for Audience Turning Points conclude, in numbers, about “Closing the Empirical Loop: a Mode Fit Every Five Minutes, a Cooling Warning and a Pre-Registered Test”?

Answer: Chapter 8 gives the five-minute fitting procedure, the Online-Curve Turning-Point Early Warner and a pre-registered test plan: a lead difference of 3 minutes needs 51 sessions and the false-alarm test needs 68. Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses.

Scope of the numbers: Marked boxes separate published evidence, runnable Python calculations (19 programs), classical sources and the author's hypotheses.

中文版

Data table

The numbers below are printed verbatim by the chapter's Python program and can be reproduced by running it. standard-library Python onlyprinted values that match the answer: 19, 51, 68

Test 1: paired lead gain, sd 7.58 min
delta 2 min: n113 sessions (23 weeks at 5 a week)
delta 3 min: n51 sessions (11 weeks at 5 a week)
delta 5 min: n19 sessions (4 weeks at 5 a week)
Test 2: premature share < 0.15 if true share is 0.055
n68 sessions; reject when at most 5 are premature
size 0.0465, power 0.8296
Show the full program
# Chapter 8. p82: how many sessions a pre-registered test needs.
# Test 1 (lead time): paired difference "DMD lead minus 12-minute-average
# lead" per session; planning sd 7.58 minutes taken from p51 output;
# two-sided alpha 0.05, power 0.8, normal approximation
# n = ((z(0.975) + z(0.8)) * sd / delta)^2 for delta = 2, 3, 5 minutes.
# Test 2 (premature warnings): one-sided exact binomial test that the
# share of warnings more than 15 minutes early is below 0.15, if its true
# value is 0.055 (p51); smallest n with power 0.8 at alpha 0.05.


import math
import statistics


z = statistics.NormalDist()
za, zb = z.inv_cdf(0.975), z.inv_cdf(0.8)
sd = 7.58
print("Test 1: paired lead gain, sd %.2f min" % sd)
for delta in (2.0, 3.0, 5.0):
    n = math.ceil(((za + zb) * sd / delta) ** 2)
    print("  delta %.0f min: n = %d sessions (%d weeks at 5 a week)"
          % (delta, n, math.ceil(n / 5)))


def cdf(k, n, p):
    return math.fsum(math.comb(n, i) * p ** i * (1 - p) ** (n - i)
                     for i in range(k + 1))


p0, p1 = 0.15, 0.055
for n in range(10, 400):
    k = -1
    while k + 1 <= n and cdf(k + 1, n, p0) <= 0.05:
        k += 1
    if k >= 0 and cdf(k, n, p1) >= 0.8:
        break
print("Test 2: premature share < %.2f if true share is %.3f" % (p0, p1))
print("  n = %d sessions; reject when at most %d are premature" % (n, k))
print("  size %.4f, power %.4f" % (cdf(k, n, p0), cdf(k, n, p1)))

Single source and citation

Fan, Y. (2026). Koopman Operators and Dynamic Mode Decomposition: Linearising Nonlinear Cash-Flow Swings and Forecasting Turning Points in Livestream Audience Curves (Version 1.0.0), Chapter 8: Closing the Empirical Loop: a Mode Fit Every Five Minutes, a Cooling Warning and a Pre-Registered Test. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-238 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026238,
  author    = {Fan, Yafu},
  title     = {Koopman Operators and Dynamic Mode Decomposition: Linearising Nonlinear Cash-Flow Swings and Forecasting Turning Points in Livestream Audience Curves},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book