Run tasks

Data analysis

Upload a dataset and run /analyze — inspection, an explicit analysis plan, cleaning with every step recorded, and a reproducible report with figures.

Start

Attach the file and describe what the data is and what you want to know:

text
/analyze Inspect the uploaded pss_trial.xlsx (two groups, PSS-10 pre/post). Check data quality and baseline balance, compute change scores, and deliver a reproducible analysis report with plots.

Say what the columns mean if the headers aren't self-explanatory — group codes, scale ranges, reverse-scored items.

What happens

  1. Inspect — structure, types, missingness, outliers, duplicates; a data-quality summary.
  2. Plan — an explicit analysis plan you can read before results: variables, transformations, tests, figures.
  3. Clean and transform — every step is code, recorded with the artifact.
  4. Analyze — the verified statistical tools run with assumption checks; effect sizes and CIs are reported alongside p-values.
  5. Interpret — a report with results, interpretation and limitations, checked by the reviewer.
The analysis turn: code steps on the left, a plot in the panel.
The analysis turn: code steps on the left, a plot in the panel.

What you get

  • <dataset>_analysis_report.md — data quality · plan · assumption checks · results · interpretation · limitations · code.
  • Plots as SVG/PNG, cleaned or derived tables as CSV, and a structured JSON of results.
The analysis report.
The analysis report.

Domain analyses

The same command drives the domain method skills — usability instruments (SUS, NASA-TLX, UEQ), semantic differential and conjoint, biosignals (ECG/HRV, EDA, respiration, EEG), face and voice from video and audio, text analysis of open-ended answers. Ask for them by name; see Method skills.

A usability A/B dashboard produced from a survey export.
A usability A/B dashboard produced from a survey export.

Walkthrough: Analyzing the data.

Experiment design PreviousStatisticsNext