NLSS

Start with your research question

NLSS helps an agent turn a question into an explicit analysis, run the calculation in R and keep the evidence. Choosing a method still depends on the design, the variables and the claim you want to make.

Explore a complete example

These workflows have been calculated with NLSS 2.0.1 and checked against direct R calculations. All study data are synthetic.

Your question See it in action What to pay attention to
How do two groups differ? Learning-study tutorial Difference, uncertainty and overlapping distributions
How should I build and compare a questionnaire score? Six items, one score Reverse scoring, missing answers, reliability and its limits
Did groups change differently over time? Change over time Linked observations, a mixed model and a planned contrast
What do additional predictors explain? Compare two models The same participants in both models, incremental fit and diagnostics
How many participants should I plan for? Power planning Meaningful effect, variability, allocation and whole participants

The examples also show data inspection, recorded transformations, figures and freely written reports. They demonstrate specific workflows; they are not a validation of every option in every method.

For another guided exploration, NLSS also includes a built-in demo with its own sample dataset. It is separate from the studies created for this website.

Find a method beyond the examples

The following capabilities are documented for version 2.0.1. All links in the table below lead directly to the NLSS GitHub repository, where each method reference describes inputs, options and limitations. These capabilities have not all been newly demonstrated on this website.

Research task Documented methods
Understand the dataset Data exploration, descriptive statistics, frequencies
Compare groups or measurements t tests, ANOVA, nonparametric tests
Investigate associations Correlations, cross-tabulations
Model an outcome Linear and generalized linear regression, linear mixed models
Examine measurement Scale analysis, reliability and agreement, exploratory factor analysis, CFA, SEM and mediation
Prepare data and handle missingness Transformations, missingness, imputation, supported pooled linear/GLM regression
Plan, inspect and communicate Power, assumptions and diagnostics, plots, reports and saved evidence

Make the decision inspectable

Start by describing the question, the study design and what each row represents. The agent can help identify missing decisions, propose an analysis and explain the result. Ask it to state the cases used, comparison direction, assumptions and uncertainty alongside the estimate.

A diagnostic flag is a reason to investigate; an unflagged result is not proof that every assumption holds. A reproducible calculation can still answer the wrong question. Keep the research decision, the numerical record and the interpretation connected.

Version 2.0.1 has explicit boundaries: linear mixed models do not imply generalized mixed models; pooling support for selected imputed regression models does not imply pooling for all methods; and a power calculation must match the intended design. Consult the linked reference before transferring an example to your own study.

Original text: Mike Hammes · CC BY 4.0. About NLSS and citation.