student-histogram

Record a sample and compute its statistical significance. Uses native-hdr-histogram and a two-tailed t-distribution table under the hood. Good for sample sizes below 200. HDR histogram is designed for "value measurements in latency and performance sensitive applications", quantizes values with a configurable precision and has a constant memory footprint.

Example

const StudentHistogram = require('student-histogram')
const h = new StudentHistogram(1, 200, 3)

// Record some example values.
const sample = [100, 120, 101, 103, 99]
sample.forEach(v => h.record(v))

const log = console.log
const percent = require('fixed-number')(1, 2, 'percent')

log('arithmetic mean            : %d ±%s', h.mean(), percent(h.rme()))
log('standard deviation         :', h.stddev())
log('minimum and maximum        : %d and %d', h.min(), h.max())
log('75% of values fall below   :', h.percentile(75))
log()

log('standard error of the mean :', h.sem())
log('margin of error            :', h.moe())
log('relative margin of error   :', h.rme())
log('with higher confidence     :', h.rme(0.998))
log()

// Estimate the size (number of sampling points aka recorded values)
// that is required to achieve a relative margin of error of 0.05.
log('recommended sample size    :', h.minimumSize(0.05))
log('for 99.5% confidence       :', h.minimumSize(0.05, 0.995))
log('with 10% error tolerance   :', h.minimumSize(0.10))

if (h.size < h.minimumSize(0.05)) {
  // Maybe record some more!
}

Output:

arithmetic mean            : 104.6 ±9.27%
standard deviation         : 7.812809993849844
minimum and maximum        : 99 and 120
75% of values fall below   : 103

standard error of the mean : 3.493994848307593
margin of error            : 9.700727296841201
relative margin of error   : 0.09274117874609179
with higher confidence     : 0.23960921458776316

recommended sample size    : 18
for 99.5% confidence       : 70
with 10% error tolerance   : 5

API

StudentHistogram(min, max, figures, opts)

Documentation to follow.

Install

With npm do:

npm install student-histogram

See also

• hdr-histogram-percentiles-obj (student-histogram is compatible)

License

MIT © 2017-present Vincent Weevers. Based in part on benchmark.js and sample-sizer. See included LICENSE file for all copyright owners.