Reliability and item-level analysis
How are reliability statistics computed?
Cronbach’s α, split-half, KMO, item-total correlations, and missing-data summaries are computed per scale on complete-case respondents. Full formulas live on the methodology page.
What does an alpha of 0.83 mean?
0.83 is in the strong band (≥ 0.80). The items in that scale are measuring one thing consistently. Read the reliability guide for thresholds, confidence intervals, and the caveats every researcher should know.
What is item-total correlation?
The correlation between an individual item and the rest of the scale (the corrected version excludes the item from the total). An item below 0.20 is a weak contributor and worth reviewing.
AI features and live results
How does AI theme extraction work?
Open-ended responses are clustered into 3 to 8 themes with counts and example quotes. Themes are generated on demand and never stored without your approval. The model never sees response IDs or any other PII.
Can I see results in real time?
Yes. Reliability statistics, item flags, and theme summaries refresh as new responses arrive. Statistics requiring a minimum sample (alpha needs at least 30 complete-case respondents) display a count instead until the threshold is met.