lognormal-distribution · shipping · also Log-normal distribution, Lognormal lifetime model

Lognormal distribution

A lognormal distribution models a positive quantity whose natural logarithm, expressed relative to a fixed unit, is normal.

Related Observatory article: Read on hesela.com

For an unshifted lifetime T > 0, ln(T/t0) is normal with mean mu and standard deviation sigma > 0; t0 is a declared time unit. The median is m = t0 exp(mu), whereas the mean is m exp(sigma^2/2). Mu and sigma describe log time, not the arithmetic mean and spread of lifetime. Changing the time unit changes mu, not sigma; the physical distribution is unchanged. [1]

For t > 0, let z = ln(t/m)/sigma. Survival is S(t) = Phi(-z) and hazard is h(t) = phi(z)/(sigma t Phi(-z)), where Phi and phi are the standard normal CDF and density. The hazard has inverse-time units and rises before eventually falling. That model shape does not establish physical rejuvenation of individual devices. [2]

Lognormal and Weibull curves can both fit a small, censored sample well yet give very different extrapolations. Select a distribution using the failure mechanism, event definition and fit diagnostics. Taking logarithms does not remove censoring: a still-working device remains an incomplete lifetime observation. A lognormal fit is not a universal law for HDD or SSD aging. [3] [1]

Relations

related
Weibull distribution, Survival function, Hazard rate

Citations

  1. NIST e-Handbook 8.1.6.4: Lognormal lifetime formulas and model scope (accessed 2026-11-18)
  2. NIST e-Handbook 1.3.6.6.9: Lognormal density, survival and hazard (accessed 2026-11-18)
  3. NIST e-Handbook 8.2.1: Choosing an appropriate life distribution model (accessed 2026-11-18)

JSON: /data/terms/lognormal-distribution.json

Published by Hesela team · Updated

Report a correction · Editorial standards · Reuse guidance

Cite this page
Hesela team. Lognormal distribution. Hesela. 2026-11-18. https://hesela.dev/terms/lognormal-distribution/