How to score Likert results: worked data and an editable spreadsheet
A ten-response teaching dataset produces eight valid scores, a mean of 3.50, a median of 4 and 62.5% agreement.
Define the item and coding first
The statement is “The setup instructions were clear.” Code strongly disagree as 1 through strongly agree as 5. Keep one not-applicable response and one missing response outside the numeric scores. These ten records are fictional examples, not results from SimilarForm customers.
Show the distribution before the average
The eight valid scores are 5, 4, 4, 3, 2, 1, 5, 4. Their sum is 28. The five agreeing or strongly agreeing responses are divided by eight valid answers, not all ten records.
| Response | Code | Count | Share of valid answers |
|---|---|---|---|
| Strongly disagree | 1 | 1 | 12.5% |
| Disagree | 2 | 1 | 12.5% |
| Neither agree nor disagree | 3 | 1 | 12.5% |
| Agree | 4 | 3 | 37.5% |
| Strongly agree | 5 | 2 | 25.0% |
| Not applicable | Excluded | 1 | Reported separately |
| Missing | Excluded | 1 | Reported separately |
Calculate and interpret
Mean = 28 ÷ 8 = 3.50. Sorted scores are 1, 2, 3, 4, 4, 4, 5, 5, so the median is 4. Top-two-box agreement = (3 + 2) ÷ 8 = 62.5%. If you coded N/A and missing as zero, the mean would fall to 2.80 for the wrong reason. The workbook exposes these formulas and updates when you edit its ten response cells.
Use the workbook
Download the XLSX and change the amber response cells using the exact labels shown in the codebook. The CSV contains the same raw example. The workbook is deliberately bounded to ten rows; extend all calculation ranges when adding respondents. Recheck the count and distribution after any edit. Unknown response labels should be corrected instead of silently treated as zero.
Reverse scoring and multi-item scales
For an intentionally reverse-worded five-point item, the reverse score is 6 minus the original score: 1 becomes 5 and 2 becomes 4. Apply it only to items whose direction is defined that way. Do not reverse N/A or missing values. Combining different questions into one score needs a defensible common construct and a missing-data rule; this single-item example does not validate a composite scale.
What 3.50 can and cannot mean
Individual response categories are ordered, but equal distances between them are an analytical assumption. Report the labels, counts, denominator and distribution alongside any mean. This tiny constructed sample supports arithmetic demonstration, not population inference, statistical significance or product-quality claims.