Rice Rule

Turkish equivalent: Rice KuralıDomain: Data Systems

A sample-size-only heuristic estimating the number of histogram bins from twice the cube root of the observation count.

Rice's rule estimates histogram bin count from sample size alone:

k ≈ 2 × n^(1/3)

Here n is the number of observations and k a count estimate. For n=125, its cube root is 5 and the result is ten bins; for n=1000, it suggests about twenty. Converting that count into actual intervals still requires the data range, a starting edge and a rounding convention.

Why the result differs from Sturges

Sturges grows as 1+log2(n), while Rice grows with the cube root. On large samples Rice commonly proposes more bins, but an increase in resolution does not by itself establish that the display is more faithful.

Rice ignores standard deviation, IQR and skewness. Two datasets with the same observation count therefore receive the same recommendation, even if one is highly concentrated and the other has long tails. Heavy ties can also make the nominal count uninformative.

Scott and Freedman–Diaconis incorporate measured spread in their width estimates. Regardless of the chosen estimator, actual interval edges must partition the observations without overlap.