Logit

Rasch & IRT · in the browser

Rasch measurement analysis in your browser

Upload a response matrix and Logit calibrates item difficulties and person abilities onto one shared logit scale — then hands you interactive Wright maps, item characteristic curves, and infit / outfit statistics. Screen for differential item functioning, convert raw scores to measures, get AI-assisted interpretation, and export publication-ready charts and tables.

Get started freeSee how it works
  • Dichotomous · RSM · PCM
  • Wright maps
  • ICC curves
  • Infit / outfit
  • DIF screening
  • Score → measure
  • AI interpretation
  • CSV · Excel · TSV · SPSS

Free to start. See pricing.

Everything between raw responses and a defensible measure

Calibrate, diagnose, and report a complete Rasch analysis — no desktop license, no fixed-width control files.

Three Rasch models
Dichotomous, Rating Scale, and Partial Credit. Estimate by joint or conditional maximum likelihood — or run both and compare them side by side.
Fit you can defend
Infit and outfit mean-squares for every item and person, with separation and reliability for the scale as a whole.
Wright maps that move
Persons and items on one logit scale — zoom into a region, hover for exact measures, export the view as SVG or PNG.
ICCs against your data
Model curves overlaid with empirical response proportions, plus category diagnostics that surface disordered thresholds.
DIF screening
Mantel–Haenszel tests across person groups catch items that behave differently for different people — before they bias the measure.
AI interpretation
A plain-language read of your calibration, fit, and DIF results, written in correct Rasch terminology. It drafts; you judge.

How it works

1

Upload responses

Drop in a CSV, Excel, TSV, or SPSS file — persons in rows, items in columns. Your data stays in your project.

2

Configure the model

Pick dichotomous, Rating Scale, or Partial Credit; estimate by JMLE or CMLE — or run both to see where they agree.

3

Read, then publish

Work through the Wright map, fit tables, ICCs, and DIF screens. Export charts and tables when the story holds up.

Validation

Checked against the published record

Every estimate Logit produces is benchmarked against published calibrations and independent engines — and those checks run as tests on every change, so agreement can’t silently drift.

r = 1.000

Winsteps · Liking for Science
The textbook Rating Scale example (Wright & Masters, 1982): all 25 published item measures reproduced, mean difference 0.04 logits.

r = 1.000

Winsteps · 2019 operational study
A five-subscale career-assessment survey reproduced against its original WINSTEPS tables — mean difference ≈ 0.01 logits.

r = 0.99999

Published MML · LSAT-6
The classic law-school admission item set versus the published marginal-maximum-likelihood calibration.

r = 1.000

Cross-engine · JMLE vs CMLE
Two estimation methods, two independent R packages, same data — item measures agree across dichotomous, RSM, and PCM.

Read the full validation dossier

Logit vs Winsteps

The de facto standard earned its place — in a different era of software. Here’s what changes when the same analysis runs in a browser.

Platform
Logit:Any browser
Winsteps:Windows only
Price
Logit:Free tier available
Winsteps:$149 – $2,250
Installation
Logit:None
Winsteps:Download + install
Wright maps
Logit:Interactive, zoomable
Winsteps:Static image
Collaboration
Logit:Share projects
Winsteps:Export files manually
Data formats
Logit:CSV, Excel, TSV, SPSS
Winsteps:Custom format
Export
Logit:SVG, PNG, CSV, JSON, PDF
Winsteps:Text files
Interface
Logit:Modern web UI
Winsteps:DOS-era interface

Frequently asked questions

What is Logit?

Logit is a browser-based workbench for Rasch measurement analysis. Upload a matrix of test or survey responses and it calibrates item difficulties and person abilities onto one shared logit scale, then gives you Wright maps, item characteristic curves, infit/outfit statistics, DIF screening, score-to-measure tables, and an exportable report.

Do I need to install anything or buy a license?

No. Logit runs entirely in the browser with a free tier — there's no Windows-only desktop program to install and no fixed-width control files to write. Upload a spreadsheet and you're analyzing.

Which Rasch models does it support?

The dichotomous model for right/wrong (0/1) data, the Rating Scale Model (RSM) for Likert-type items that share one response scale, and the Partial Credit Model (PCM) for items with item-specific step structure. Estimation is by joint maximum likelihood (JMLE, via TAM) or conditional maximum likelihood (CMLE, via eRm) — and you can run both and compare.

What data formats can I upload?

CSV, TSV, Excel (.xlsx and .xls), and SPSS (.sav) — persons in rows, items in columns. SPSS files keep their variable names and value labels.

Can I trust the numbers?

Logit's estimates are produced by the same R engines the research literature uses (TAM and eRm) and are checked continuously against published reference calibrations and independent estimators — including a Winsteps-published textbook dataset and a classic MML-calibrated item set — with those checks run as automated tests on every change. See the Validation page for the current benchmarks and their agreement.

Is it a Winsteps replacement?

For the analyses most researchers run day to day — calibration, Wright maps, fit, category structure, DIF, score-to-measure — yes, and the estimates are checked head-to-head against Winsteps output. Some advanced Winsteps-only features (for example residual-based dimensionality analysis) are still on the roadmap; the Validation page is candid about what is and isn't covered yet.

Is my data private?

Your uploads live inside your own project, scoped to your account — other users can't see them. You can delete a dataset or analysis at any time.

What isn't supported yet?

Residual-based dimensionality diagnostics (principal-components analysis of residuals) and some of Winsteps' more specialized tables are on the roadmap. New psychometric features don't ship without a reference check or a simulation-recovery study behind them.

Still curious how the method works? Start learning Rasch analysis.

From response matrix to defensible measure in one sitting

Upload a dataset and see your first Wright map in minutes.

Get started free