Understand: calculate and experiment without data of your own
In brief
The “Understand” tab in the top bar needs no project and no data. It recalculates what appears in lessons, textbooks or a study, and shows it as a picture with a number and a sentence. Nothing is saved and nothing ends up in a .statable file; the inputs stay only while the page is open.
The five tiles
- Calculate distributions: probabilities, quantiles, mean and spread for the normal, t, χ², F, binomial, Poisson and hypergeometric distributions and for your own table.
- Calculate from summary numbers: confidence interval and test for a proportion or a mean when only the numbers are available, and the sample size for a desired width.
- Understand the test for a proportion: the binomial test with rejection region, actual α, p value, type II error and prediction interval.
- Learning lab: sampling distribution, bootstrap, randomisation, random experiments, four-field table and the coverage of confidence intervals.
- Allocate by chance: select people from a list or distribute them to groups (blocks, strata, visible seed), as a teaching tool.
Limits
- Binomial distribution up to n = 1,000,000, Poisson distribution up to λ = 100,000, hypergeometric distribution up to N = 10,000,000. Above that Statable shows a note instead of an inaccurate number.
- Very small probabilities are never shown as “0” but in exponent notation (for example 1.2 · 10⁻⁹).
For specialists
Everything is calculated in the browser in JavaScript, without PSPP and without R. The distributions are checked against R (pnorm, pt, pchisq, pf, pbinom, ppois, phyper and the q functions) and against PSPP (CDF and IDF functions), to a relative 1e-9 for probabilities from 1e-12. Upper tails are calculated directly, never as 1 − P.