larzstats

Statistics without numpy/scipy. Pure Python, zero dependencies.

The everyday statistics you reach for on a list of numbers — central tendency and spread, quantiles, correlation, a least-squares fit you can predict from, and the normal distribution — all from math alone, working back to Python 3.8 (where the stdlib statistics module still lacks regression and correlation).

from larzstats import describe, linregress, correlation, normal_cdf

describe([2, 4, 4, 4, 5, 5, 7, 9])
# {'count': 8, 'mean': 5.0, 'std': 2.13.., 'min': 2, 'q1': 4.0,
#  'median': 4.5, 'q3': 5.5, 'max': 9}

fit = linregress([1, 2, 3, 4], [2.1, 3.9, 6.1, 8.0])
fit.slope, fit.r2, fit.predict(5)

correlation([1, 2, 3], [2, 4, 6])      # 1.0
normal_cdf(1.96)                       # 0.975

Why

• The functions the stdlib doesn't give you (on 3.8). linregress (with slope/intercept/r/r2 and predict()), correlation, covariance, describe, iqr, z_score, normal_cdf/normal_pdf — no statistics version gate, no numpy.
• Complete descriptives. mean/median/mode, geometric & harmonic means, sample and population variance/stdev, quantiles/percentiles/quartiles.
• Zero dependencies. For scripts, dashboards, tests, and teaching — anywhere numpy is overkill. Pairs with larzchart to plot the results.

Install

pip install larzstats

Usage

from larzstats import (mean, median, mode, variance, stdev, quantile, percentile,
                       iqr, describe, correlation, linregress, normal_cdf, z_score)

mean(xs); median(xs); stdev(xs)            # sample by default; sample=False for population
quantile(xs, 0.9); percentile(xs, 90); iqr(xs)
describe(xs)                               # summary dict

fit = linregress(x, y); fit.predict(x0); fit.r2
correlation(x, y)                          # -1..1
normal_cdf(x, mu=0, sigma=1); z_score(x, mu, sigma)

Tests

python -m unittest discover -s tests -v   # 20 tests incl. regression + normal CDF

The Larz stack

One of 30+ pure-Python, zero-dependency libraries at github.com/larz-scripter.

License

MIT © larz-scripter