LoginSign Up
← Back to all sparks
S

statpsych

INFRA · APIS
Velocity0.0

Statistical Methods for Psychologists

github.com/dgbonett/statpsych ↗

A statistics catalogue for psychology that grows by the release and rarely changes shape.

psychometricsconfidence-intervalssample-sizestatistical-powerr-package
◆Current state
statpsych supplies confidence intervals, hypothesis tests, power calculations and sample-size planning for the designs psychology researchers actually run, exposed as several hundred small named functions rather than a modelling framework. Version 2.0.0 adds eight functions across logistic model performance, Kendall tau-a intervals and sample sizes, intraclass correlation testing, Geary kurtosis and Mann-Whitney power, and retires three names in favour of generalised replacements. The major version number reflects those removals rather than a change in how the package is used.
◆Where it's heading
Every release in this window is the same shape: a list of new functions, occasionally a rename. The package grows by filling cells in a grid of estimand, design and inferential goal, and 2.0.0 is notable only for finally deleting the three names its generalised replacements had superseded. That makes it a reference library whose value is coverage and stability, not direction, and the entries give no sign of that changing.
◆Prediction
Expect the accretion to continue along the same axes, with sample-size and power counterparts filled in for estimands that currently have interval functions but no planning ones. The 2.0.0 deletions suggest occasional consolidation passes when a generalised function makes older specific ones redundant.

◆Recent moves

  1. 1mo ago

    Kendall tau, logistic fit measures added; three names retired

    Adds eight functions spanning logistic model performance, Kendall tau-a intervals and sample sizes, intraclass correlation testing, Geary kurtosis and Mann-Whitney power, and removes iqv and two slope-specific sample-size functions in favour of generalised replacements. The deletions are what earn the major version; the additions continue the catalogue's usual growth.

    View source ↗
  2. 7mo ago

    Intraclass correlation and diversity indices gain full coverage

    Ten new functions, with intraclass correlation getting the complete treatment of interval, sample size and prior-informed sample size, alongside diversity indices, generalised means, standard deviation intervals and Scheffe contrasts for means and proportions. A representative release for this package: breadth across designs rather than depth in any one.

    View source ↗
  3. 1y ago

    statpsych 1.8

    A version tag whose notes contain only a merge of a development branch and a DESCRIPTION update. Nothing in the entry indicates what, if anything, changed for users.

    View source ↗
  4. 2y ago

    Correlation tests and finite-population corrections added

    Extends the catalogue into finite population corrections for means and proportions, Poisson rates and their ratios, biserial correlation, prediction intervals, and a run of Pearson and Spearman hypothesis tests supporting non-zero null values. The non-zero null support is the more useful detail, since it covers equivalence-style questions the default tests cannot express.

    View source ↗
  5. 2y ago

    Coefficient of variation and 2x2 within-subjects effects

    Adds coefficient-of-variation intervals and ratios, predictive values under retrospective sampling, and standardised and median-based effects for 2x2 within-subjects and mixed designs. The 2x2 additions fill in design cells the package already covered for between-subjects layouts.

    View source ↗
  6. 3y ago

    Power calculations added for means, proportions and correlations

    Introduces the power family across one-sample, two-sample and paired tests of means and proportions, linear contrasts and correlations, giving the package a planning counterpart to the intervals it already computed. This is the release that establishes power as a standing axis the later versions keep extending.

    View source ↗