Cross-Tab — Reading Significance Letters
A cross-tab counts how two label columns intersect: defect type by shift, reject reason by supplier, answer by department. The percentages alone are not enough — a shift with 30 parts inspected and one with 300 will bounce around differently by pure chance.
That is what the letters are for. Each column gets a letter (A, B, C…), and a cell carries the letters of every column it is significantly higher than. So "39.5% ACD" in the Maintenance column reads as: Maintenance is significantly higher than columns A, C and D. No letter means the difference is within what chance could produce.
Two things to check before you believe a table: the base (n) under each column heading — below 30 the test is weak — and the chi-square line at the top, which asks "is there any difference at all?" before you start reading individual cells. Keep Bonferroni on: with 4 columns you are making 6 comparisons per row, and without the correction a few letters will appear by chance alone.
That is what the letters are for. Each column gets a letter (A, B, C…), and a cell carries the letters of every column it is significantly higher than. So "39.5% ACD" in the Maintenance column reads as: Maintenance is significantly higher than columns A, C and D. No letter means the difference is within what chance could produce.
Two things to check before you believe a table: the base (n) under each column heading — below 30 the test is weak — and the chi-square line at the top, which asks "is there any difference at all?" before you start reading individual cells. Keep Bonferroni on: with 4 columns you are making 6 comparisons per row, and without the correction a few letters will appear by chance alone.
Try it in the app
Try: Load "Employee Survey — Engagement" → Analyses → Cross-Tab → rows = Main_Issue, columns = Department. Maintenance carries letters on the Equipment row. Now switch Bonferroni off and watch extra letters appear — those are the ones you should not trust.