Descriptive statistics

Overview

Multik provides three core descriptive statistics functions through mk.stat:

Function	Description
`mean`	Arithmetic mean of all elements, or along a given axis.
`median`	Median (middle value) of all elements.
`average`	Weighted average over all elements.

mean

Returns the arithmetic mean of all elements in the array, or along a given axis.

Signatures

// Scalar mean (all elements)
fun <T : Number, D : Dimension> Statistics.mean(a: MultiArray<T, D>): Double

// Mean along axis
fun <T : Number, D : Dimension, O : Dimension> Statistics.mean(
    a: MultiArray<T, D>,
    axis: Int
): NDArray<Double, O>

Dimension-specific overloads are also available:

fun <T : Number> Statistics.meanD2(a: MultiArray<T, D2>, axis: Int): NDArray<Double, D1>
fun <T : Number> Statistics.meanD3(a: MultiArray<T, D3>, axis: Int): NDArray<Double, D2>
fun <T : Number> Statistics.meanD4(a: MultiArray<T, D4>, axis: Int): NDArray<Double, D3>
fun <T : Number> Statistics.meanDN(a: MultiArray<T, DN>, axis: Int): NDArray<Double, DN>

Parameters

Parameter	Type	Description
`a`	`MultiArray<T, D>`	Input array.
`axis`	`Int`	Axis along which to compute mean.

Returns: Double for the scalar form, or NDArray<Double, O> for the axis form.

Example

val a = mk.ndarray(mk[mk[1, 2, 3], mk[4, 5, 6]])

mk.stat.mean(a)                                  // 3.5
mk.stat.meanD2(a, axis = 0)                      // [2.5, 3.5, 4.5]
mk.stat.meanD2(a, axis = 1)                      // [2.0, 5.0]

median

Returns the median of all elements in the array. Elements are sorted and the middle value is returned (or the average of the two middle values for even-length arrays).

Signature

fun <T : Number, D : Dimension> Statistics.median(a: MultiArray<T, D>): Double?

Parameters

Parameter	Type	Description
`a`	`MultiArray<T, D>`	Input array.

Returns: Double? — the median value, or null if the array is empty.

Example

val a = mk.ndarray(mk[3, 1, 4, 1, 5, 9, 2, 6])

mk.stat.median(a)   // 3.5  (average of 3 and 4 after sorting)

val b = mk.ndarray(mk[1, 2, 3])
mk.stat.median(b)   // 2.0

average

Returns the weighted average of all elements. When no weights are provided, all elements are weighted equally (equivalent to mean).

Signature

fun <T : Number, D : Dimension> Statistics.average(
    a: MultiArray<T, D>,
    weights: MultiArray<T, D>? = null
): Double

Parameters

Parameter	Type	Description
`a`	`MultiArray<T, D>`	Input array.
`weights`	`MultiArray<T, D>?`	Optional weights array (same shape as `a`). Default: `null`.

Returns: Double — the weighted average.

Example

val values = mk.ndarray(mk[1.0, 2.0, 3.0, 4.0])

// Unweighted (equivalent to mean)
mk.stat.average(values)                                      // 2.5

// Weighted
val weights = mk.ndarray(mk[1.0, 1.0, 1.0, 5.0])
mk.stat.average(values, weights)                             // 3.25
// Calculation: (1*1 + 2*1 + 3*1 + 4*5) / (1+1+1+5) = 26/8 = 3.25

Pitfalls

mean always returns Double, even for integer arrays.
median returns Double? — it can be null for empty arrays.
Axis-based mean overloads require explicit dimension type parameters, e.g. mk.stat.meanD2(a, axis = 0).
The weights array in average must have the same shape as the input array.

22 February 2026