How to use it
- Put each respondent on its own row and separate rows with semicolons.
- In each row, list that person’s scores for every item in the same order, separated by commas or spaces.
- Read alpha, its label and alpha if each item is deleted.
Formula
α = k ÷ (k − 1) × (1 − Σ sᵢ² ÷ s_T²)
- k is the number of items,
- sᵢ² is the variance of item i across respondents,
- s_T² is the variance of the respondents’ total scores.
Variances use n − 1 in the denominator; the ratio is the same either way. Alpha if item deleted applies the same formula to the remaining k − 1 items (shown when there are at least 3 items).
Worked example
Three respondents answer two items: (2, 3), (4, 4), (3, 5).
Item 1 scores 2, 4, 3 have variance 1; item 2 scores 3, 4, 5 have variance 1. The totals 5, 8, 8 have mean 7 and variance (4 + 1 + 1) ÷ 2 = 3. α = 2 ÷ 1 × (1 − 2/3) = 0.67, questionable by the rule of thumb.
Tips and common mistakes
- Alpha rises with the number of items, so a long scale can reach a high alpha even when the items are only loosely related.
- Alpha assumes the items measure one thing. For a questionnaire with several subscales, compute alpha for each subscale.
- Leave out respondents with missing answers or fill them in first: every row must have the same number of scores.