What does Error Bars in Dating conclude, in numbers, about “A Date Is a Distribution: Highest-Density Ranges, Consistency Tests and Prior Windows”?

Answer: A distribution with several peaks cannot be summarised by one number. The 68.3 % and 95.4 % levels can give different numbers of segments, and on a plateau the mean, median and mode can lie a century apart. Before several measurements are pooled they must pass a consistency test, or one outlier slips into the result wearing a smaller error. What genuinely narrows a range is an independent prior unrelated to the claim under test, and a stylistic judgement cannot play that role. Here the story of radiocarbon pauses. The next chapter turns to the clock that porcelain and pottery actually rely on, which measures not carbon but the radiation dose that mineral grains have accumulated since they were last heated.

Scope of the numbers: Objects, prices and parameters are illustrative, the calibration curve is a hand-set illustrative line rather than IntCal data, and all calculations are generated by the book's programs. Chapter 6 writes testing as a problem of expected loss; under illustrative parameters a buyer should not buy untested below a prior of 0.655, and a test is worth most at middling priors.

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: 68.3, 95.4

68.3% HPD: 5 range(s)
794 to744 BCmass 0.128
720 to690 BCmass 0.077
661 to607 BCmass 0.137
595 to560 BCmass 0.090
509 to410 BCmass 0.252
95.4% HPD: 2 range(s)
803 to541 BCmass 0.627
538 to403 BCmass 0.328
summaryyear BCdensity / peak
mean6030.896
median6050.908
mode7051.000
sd of curve BP over 780 to 410 BC: 10.4
Show the full program
# Chapter 3. p31: a calibrated date is a distribution, not a year.
# Same SCHEMATIC curve as p21 (not IntCal). For 2450 +/- 25 BP the
# 68.3% and 95.4% highest-density ranges are listed with their mass,
# and three one-number summaries (mean, median, mode) are compared
# with the density at each of them, relative to the peak density.
import math
import statistics

KNOTS = ((1000, 2830), (820, 2500), (780, 2440), (730, 2470),
         (680, 2430), (600, 2465), (540, 2425), (470, 2455),
         (410, 2440), (360, 2340), (200, 2160))
CURVE_SD = 15.0


def curve(bc):
    for (x0, y0), (x1, y1) in zip(KNOTS, KNOTS[1:]):
        if x1 <= bc <= x0:
            return y0 + (y1 - y0) * (x0 - bc) / (x0 - x1)
    raise ValueError(bc)


def posterior(m, s, csd=CURVE_SD):
    years = list(range(1000, 199, -1))
    v = s * s + csd * csd
    w = [math.exp(-(m - curve(y)) ** 2 / (2.0 * v)) for y in years]
    tot = math.fsum(w)
    return years, [x / tot for x in w]


def hpd(years, p, level):
    order = sorted(range(len(p)), key=lambda i: (-p[i], i))
    keep, acc = set(), []
    for i in order:
        keep.add(i)
        acc.append(p[i])
        if math.fsum(acc) >= level:
            break
    segs, start = [], None
    for i in range(len(p) + 1):
        inside = i < len(p) and i in keep
        if inside and start is None:
            start = i
        if not inside and start is not None:
            mass = math.fsum(p[start:i])
            segs.append((years[start], years[i - 1], mass))
            start = None
    return segs


def quantile(years, p, q):
    acc = []
    for y, x in zip(years, p):
        acc.append(x)
        if math.fsum(acc) >= q:
            return y
    return years[-1]


if __name__ == "__main__":
    ys, p = posterior(2450, 25)
    for lev in (0.683, 0.954):
        segs = hpd(ys, p, lev)
        print("{:.1%} HPD: {} range(s)".format(lev, len(segs)))
        for a, b, mass in segs:
            print("   {:>4} to {:>4} BC  mass {:.3f}".format(a, b, mass))
    mean = math.fsum([y * x for y, x in zip(ys, p)])
    med = quantile(ys, p, 0.5)
    top = max(p)
    mode = ys[p.index(top)]
    print("summary  year BC  density / peak")
    for name, y in (("mean", round(mean)), ("median", med),
                    ("mode", mode)):
        d = p[ys.index(y)] / top
        print("  {:<6}  {:>7}  {:.3f}".format(name, y, d))
    spread = statistics.pstdev([curve(y) for y in range(780, 409, -1)])
    print("sd of curve BP over 780 to 410 BC: {:.1f}".format(spread))

Single source and citation

Fan, Y. (2026). Error Bars in Thermoluminescence and Radiometric Dating: Multiple Solutions on Calibration Plateaus and When to Send Antiques for Testing (Version 1.0.0), Chapter 3: A Date Is a Distribution: Highest-Density Ranges, Consistency Tests and Prior Windows. 5D Library. DOI pending

DOI pending · monograph page goes live after publication · 5DL-B-331 · v1.0.0 · CC-BY-NC-ND-4.0

BibTeX
@book{fan2026331,
  author    = {Fan, Yafu},
  title     = {Error Bars in Thermoluminescence and Radiometric Dating: Multiple Solutions on Calibration Plateaus and When to Send Antiques for Testing},
  year      = {2026},
  publisher = {5D Library},
  edition   = {1.0.0},
  note      = {DOI pending}
}

Other chapter answers in this book