The spectral test is a statistical test for the quality of a class of pseudorandom number generators (PRNGs), the linear congruential generators (LCGs).
[1] LCGs have a property that when plotted in 2 or more dimensions, lines or hyperplanes will form, on which all possible outputs can be found.
[2] The spectral test compares the distance between these planes; the further apart they are, the worse the generator is.
[3] As this test is devised to study the lattice structures of LCGs, it can not be applied to other families of PRNGs.
According to Donald Knuth,[4] this is by far the most powerful test known, because it can fail LCGs which pass most statistical tests.
The IBM subroutine RANDU[5][6] LCG fails in this test for 3 dimensions and above.
Let the PRNG generate a sequence
be the maximal separation between covering parallel planes of the sequence
The spectral test checks that the sequence
Knuth recommends checking that each of the following 5 numbers is larger than 0.01.
is the modulus of the LCG.
Knuth defines a figure of merit, which describes how close the separation
Under Steele & Vigna's re-notation, for a dimension
is the Hermite constant of dimension d.
is the smallest possible interplane separation.
[7]: 3 L'Ecuyer 1991 further introduces two measures corresponding to the minimum of
across a number of dimensions.
for a LCG from dimensions 2 to
is the same for a multiplicative congruential pseudorandom number generator (MCG), i.e. one where only multiplication is used, or
Steele & Vigna note that the
is calculated differently in these two cases, necessitating separate values.
[7]: 13 They further define a "harmonic" weighted average figure of merit,
[7]: 13 A small variant of the infamous RANDU, with
has:[4]: (Table 1) The aggregate figures of merit are:
[a] George Marsaglia (1972) considers
as "a candidate for the best of all multipliers" because it is easy to remember, and has particularly large spectral test numbers.
[9] The aggregate figures of merit are:
[a] Steele & Vigna (2020) provide the multipliers with the highest aggregate figures of merit for many choices of m = 2n and a given bit-length of a.
They also provide the individual
values and a software package for calculating these values.