Column letters, not p-values
Free crosstab significance testing for market researchers. Paste a banner table from Excel and get pairwise z-test results in seconds.
What StatLetter does
StatLetter is a free, browser-based significance testing tool for survey crosstabs. It runs pooled two-proportion z-tests across banner columns and outputs standard significance letters (UPPERCASE for 95% confidence, lowercase for 90%) directly in your browser. Calculations execute locally in client-side JavaScript; table data, labels, base sizes, and results are never sent to a server.
Frequently asked questions
What statistical test does StatLetter use for significance testing?
StatLetter runs an unweighted, pooled two-proportion z-test on pairwise banner columns. The pooled proportion is calculated across both comparison groups to determine the standard error. Tests are two-tailed, evaluating whether the observed percentage in one column is statistically higher than the percentage in another column at the selected confidence threshold.
Why does StatLetter use a pooled z-test for crosstab columns?
When testing independent subgroups (such as demographic segments or geographic regions), the null hypothesis assumes that both groups share the same underlying population proportion. Pooling the observed data across both columns provides the best estimate of that shared proportion to calculate the standard error. StatLetter uses this pooled variance approach for all pairwise proportion tests.
How are column significance letters assigned in the table?
StatLetter uses the lettering convention common in market research banner tables. Each column is assigned an identifying letter (A, B, C, etc.). When Column A has a proportion significantly higher than Column B, the letter "B" appears in Column A's cell. If dual-level testing is enabled, uppercase letters denote significance at the higher confidence level (typically 95%), and lowercase letters denote significance at the lower level (typically 90%).
Does StatLetter support significance testing for means or rating scales?
No. StatLetter currently evaluates proportion and percentage data only. It does not run independent-sample t-tests on rating-scale means, standard deviations, or numeric averages. If your crosstabs require means testing, you will need to use dedicated tabulation software or statistical packages.
Can StatLetter account for survey weights or effective base sizes?
No. StatLetter tests raw entered percentages against the unweighted base sizes provided in the base row. It does not calculate design effects, weighting efficiency, or effective base sizes. If you are working with weighted data, you must calculate effective base sizes prior to entry or use a full-scale tabulation engine.
How does StatLetter handle data privacy?
All calculations run locally inside your browser using client-side JavaScript. Table contents, column labels, base sizes, and test outputs are never transmitted to an external server, logged in a database, or stored in cookies. Like all websites, loading the web page sends a standard HTTP request to the web host, but your survey data remains entirely on your machine.