Split-half reliability
Purpose
Assess consistency among items intended to measure the same construct.
Data and assumptions
Reverse-score relevant items first. Each scale uses cases complete on its items. High reliability alone does not establish construct validity.
Configure this method
Split by odd/even or first/last selected-item positions. Results include the half-score correlation, Spearman–Brown and Guttman coefficients; unequal halves use the unequal-length correction.
Variables
| Control | Input |
|---|---|
| Analysis variables | Select variables |
| Dimensions and items | Add named dimensions and their items |
Analysis settings
| Parameter | Choices | Default |
|---|---|---|
| Also analyze the total scale | Set as needed | Disabled |
| Split method | Odd and even items, First and second halves | odd_even |
Analysis parameters remain visible in the main configuration area. Advanced formatting contains only the table title and numeric decimals. Methods that calculate intervals also show the significance level; 0.05 corresponds to 95% intervals.
Read the results
Consider coefficients together with item count and content. Negative coefficients call for checking coding and item direction. Item-deleted estimates inform judgment rather than automatic deletion.
Workflow and exports
- Upload the dataset and choose the method from the sidebar or search.
- Select the variables above, set the design-specific options, then run the analysis.
- Inspect the paper-style result. Reconfigure from the result area, or open raw output for diagnostics and sample details.
- Rename and keep result tabs as needed. Export Word, Excel, CSV or TXT; batch export can include the selected AI interpretation. Word retains figures. AI text should be checked against the estimates.