Array Summary

Shipping

array_summary

Twelve descriptive statistics of any array in one pass: mean, spread, shape, extrema, count

Signature

Inputs

dataArrayrequired— Any Vector / Matrix / Tensor; flattened to 1-D in C (row-major) order before reduction.

Outputs

meanScalar— Arithmetic mean $\bar x$.
geometric_meanScalar— Geometric mean $\left(\prod x_i\right)^{1/n}$ (defined for positive data).
harmonic_meanScalar— Harmonic mean $n / \sum (1/x_i)$.
varianceScalar— Variance $\sigma^2$.
std_devScalar— Standard deviation $\sigma$.
skewnessScalar— Third standardized moment (asymmetry).
kurtosisScalar— Fourth standardized moment (tailedness).
minScalar
maxScalar
sumScalar
medianScalar— 50th percentile.
countInteger— Number of elements $n$.

Description

Array Summary computes a complete descriptive-statistics profile of an array in a single evaluation and emits all twelve results on separate ports — no information is dropped and no second pass is required. The input may be a Vector, Matrix, or Tensor; it is flattened to a 1-D sample in C (row-major) order, so shape is irrelevant to the reduction.

The outputs cover the three classical families: location (arithmetic, geometric, harmonic mean, median), spread (variance, standard deviation, min, max, sum), and distribution shape (skewness, kurtosis), plus the sample count. Because every statistic is exposed as its own scalar port, you wire only the ones you need downstream while the rest are computed for free.

The node is stateless: the output depends solely on the current input array, with no memory across evaluations.

Mathematics

\overset{x}{ˉ} = \frac{1}{n} i = 1 \sum n x_{i}, σ^{2} = \frac{1}{n} i = 1 \sum n (x_{i} - \overset{x}{ˉ})^{2}

skew = \frac{1}{n} \sum (\frac{x _{i} - x ˉ}{σ})^{3}, kurt = \frac{1}{n} \sum (\frac{x _{i} - x ˉ}{σ})^{4}

GM = (i = 1 \prod n x_{i})^{1/ n}, HM = \frac{n}{\sum _{i = 1}^{n} 1/ x _{i}}

Examples

Profiling a measurement batch

Feed a Vector of 5 samples $[1, 2, 3, 4, 5]$ . Outputs: mean $= 3$ , median $= 3$ , sum $= 15$ , min $= 1$ , max $= 5$ , count $= 5$ . The std_dev, skewness, and kurtosis ports characterize the distribution's spread and shape in the same shot.

Applications

One-glance characterization of a sensor buffer or data column before deeper analysis.
Feeding location/spread scalars into thresholds, gauges, or downstream normalization.
Comparing distribution shape (skewness/kurtosis) across experimental runs.

Neat

Backed by the ndarray-stats surface: a single call returns all twelve statistics, and the executor forwards each on a dedicated port in a locked order that mirrors the underlying summary struct.

Any-rank input is accepted — a Matrix or Tensor is flattened C-order, so the same node profiles a scalar buffer or an image tensor identically.