Geometric Mean

Computes the geometric mean of a numeric vector, with optional weights. Useful for aggregating multiplicative quantities such as ratio-based health indicators (e.g. UHC service-coverage tracers, where the composite index is the geometric mean of component coverage values).

Usage

geomean(x, w = NULL, na.rm = TRUE)

Arguments

x

A numeric vector. Zeros produce a result of 0. Negative values produce NaN with a warning, since the geometric mean is undefined for negative numbers.

w

Optional numeric vector of weights, the same length as x. Must be non-negative. If NULL (default), the unweighted geometric mean is returned.

na.rm

Logical. Should missing values in x (and w, if provided) be removed before computation? Default TRUE.

Value

A numeric scalar. Returns NA_real_ when the input is empty, when it is entirely NA, or when na.rm = FALSE and any element is NA. Returns NaN with a warning when x contains negative values, or when all weights are zero.

Details

Pass w to compute a weighted geometric mean, defined as exp(weighted.mean(log(x), w)).

Examples

# Unweighted
geomean(c(1, 4, 16))                # 4
#> [1] 4
geomean(c(0.6, 0.8, 0.95))          # ~0.772 — typical UHC tracer aggregation
#> [1] 0.7697002
geomean(c(1, NA, 4))                # 2
#> [1] 2
geomean(c(1, NA, 4), na.rm = FALSE) # NA_real_
#> [1] NA
geomean(c(1, 0, 4))                 # 0
#> [1] 0

# Weighted
geomean(c(1, 4, 16), w = c(1, 1, 1))     # 4 (equal weights = unweighted)
#> [1] 4
geomean(c(1, 4, 16), w = c(1, 2, 1))     # weighted toward 4
#> [1] 4
geomean(c(0.6, 0.8, 0.95), w = c(2, 1, 1))
#> [1] 0.7232343