Statistics Glossary → Geometric and harmonic mean
Geometric and harmonic meanconcept
Two kinds of average used instead of the arithmetic mean for positive data that grow multiplicatively or are expressed as rates.
The geometric mean is the n-th root of the product of n positive values; equivalently, it is the exponential of the mean of their logarithms: GM = exp(Σ ln xᵢ / n). It suits growth rates, interest, bacterial counts and right-skewed, roughly log-normal data. The harmonic mean is the reciprocal of the mean of the reciprocals: HM = n / Σ(1/xᵢ); it gives the correct average for quantities such as speeds or rates per unit.
For positive values the ordering HM ≤ GM ≤ arithmetic mean always holds, with equality only when all values are the same. Both are defined only for values greater than zero: if one value is 0 the geometric mean becomes 0 or cannot be computed, and with negative values it loses its meaning. When an analysis is done on the log scale, the back-transformed mean is the geometric mean, not the arithmetic one.
An everyday example
An investment gains 50% in the first year and loses 50% in the second. The growth factors are 1.5 and 0.5, whose arithmetic mean is 1.0, suggesting 'no change on average'. Yet 100 has become 150 and then 75. The geometric mean is √(1.5 × 0.5) ≈ 0.866, so the investment actually shrank by about 13.4% per year.
In MerQur
In Descriptive Statistics, the 'CV, geometric/harmonic mean, sum' option gives the geometric mean as an extra row together with the coefficient of variation, the harmonic mean and the sum; if the variable has zero or negative values, the geometric mean is left blank.