Now live · Analysis workbench
A statistical workbench that runs meta-analysis, confirmatory factor analysis and structural equation modelling in your browser — on the same R packages your reviewers already trust. It explains every number it gives you, and it writes up the methods section while you work.
No sign-up. Nothing to install. Works on a laptop in a browser tab.
What it does
Each one is a complete workflow — from a spreadsheet to a figure and a paragraph you can defend.
Pool results across studies and find out why they disagree. Every effect measure a review needs — risk and odds ratios, mean differences, correlations, proportions and rates, or your own pre-computed estimates.
Test whether your questionnaire measures what you say it measures. Specify which items belong to which construct and see whether the data agree.
Draw the paths between your variables — measured or latent — and fit them all at once. Built around the question most people actually came for: how much of the effect runs through the mediator?
Exploratory factor analysis and scale reliability are next on the roadmap.
How it works
The workbench is built around the order you actually work in, not the order a statistics textbook is written in.
Upload a CSV, paste straight from Excel or SPSS output, or open one of the worked examples. Columns do not have to be named anything in particular.
Say what your columns mean and choose your model. Sensible defaults are already set, with every option explained where you meet it — not in a manual.
Statistics, diagnostics and figures, each with a sentence saying what it means and what it does not. Download any figure as SVG or PNG and any table as CSV.
A methods and results section in past tense, with your numbers in it, ready to paste into a manuscript and edit. Export the whole report to Word.
Why you can trust the numbers
A tool that is easy to use and quietly wrong is worse than no tool at all. So the workbench does not compute the statistics itself — it drives the established R packages and shows you what they returned.
Thirteen real datasets ship with the workbench, exported unaltered from the R packages that publish them — BCG vaccine trials, Holzinger & Swineford's ability tests, Bollen's study of industrialisation and democracy, and more. Nothing is simulated. When you are learning a method, you want to be able to check your answer against the literature.
A project lives in your own browser storage, so you can close the tab and come back to it. Nothing is uploaded until you press run, and then only the columns the model needs are sent — used to compute your result and not stored afterwards. There is no account, so there is nothing to sign in to and nothing to leak.
Who it is for
Run the analysis your supervisor asked for, understand what it says, and hand in a methods section that names the software and version.
Pool your extracted studies, investigate heterogeneity properly, and export a forest plot at publication resolution.
A second opinion on a model, or a fast way to check an analysis before it goes to a client — without a licence for software you use twice a year.
The workbench does the computing. If you would rather have a statistician alongside you — designing the analysis, interpreting the output, or defending it to a reviewer — that is what our research consulting service is for.
Bring a spreadsheet, or start from one of the worked examples. It takes about a minute to get your first result.