Supported Methods Matrix

Current support snapshot for statistical methods, model families, and analysis capabilities in the current Licklider release.

This page is the authoritative release-facing reference for which statistical methods are currently supported in Licklider.

For the rationale behind particular defaults, see JS vs Python Execution Model.

Statistical methods

MethodStatusNotes
Welch's t-testSupportedDefault for independent two-group comparisons
Student's t-testSupportedAvailable when the analysis is intentionally configured that way
Mann-Whitney USupportedUsed as a non-parametric alternative for independent groups
Paired t-testSupportedAvailable for paired or repeated observations with two conditions
Wilcoxon signed-rankSupportedNon-parametric paired alternative
One-way ANOVASupportedStandard multi-group comparison workflow
Kruskal-WallisSupportedNon-parametric multi-group alternative
Two-way ANOVASupportedDefault engine policy uses Type II sum of squares
Repeated measures ANOVASupportedAvailable when the repeated structure is recognized and the workflow fits
FriedmanSupportedNon-parametric repeated-measures alternative
Linear regression (OLS)SupportedFor continuous outcome regression within supported assumptions
Logistic regressionSupportedBinary outcome modeling support is present
Generalized linear mixed modelsSupportedCurrent support is focused rather than universal across every mixed-model family
Kaplan-MeierSupportedSurvival analysis support is part of the current public surface
Cox proportional hazardsSupportedAvailable within the current survival-analysis scope
Non-linear regression (4PL / Hill)SupportedAvailable for supported dose-response style workflows
Chi-square testSupportedCurrent categorical association support includes chi-square workflows
Fisher's exact testNot yet supported as a first-class public surfaceDo not assume parity with chi-square support yet
Bootstrap confidence intervalsSupportedIncluded as part of robustness and estimation workflows
Permutation testsSupportedAvailable as a robustness-oriented alternative
Bayes factor supplementSupportedTreated as a supplement rather than a replacement for the main frequentist result
Power analysisPartialCurrent public support is narrower than the full method surface

How to read "supported"

"Supported" on this page means the current release exposes the method as part of the intended product experience.

It does not mean:

  • every adjacent workflow variant is equally mature
  • every edge case is automatically recognized
  • every possible import shape or design flaw is handled without researcher input

Power analysis scope

Power analysis should currently be read conservatively.

  • Core coverage is strongest for common t-test, ANOVA, and chi-square style use cases.
  • Coverage is narrower for regression, mixed-model, and survival-analysis planning.
  • If your planning question depends on a highly specialized design, do not assume first-class support without checking the surrounding docs and implementation details.

Data and workflow assumptions

Method availability depends not only on the statistical engine, but also on the recognized data shape and workflow context.

Examples:

  • paired methods depend on a usable paired or subject-like structure
  • repeated-measures paths depend on the data encoding repeated observations in a recognizable way
  • regression support depends on a valid distinction between outcome, predictors, and design context