Cronbach’s alpha
Internal consistency of a set of items that should measure one thing.
When to use it
Use Cronbach’s alpha when several items are supposed to measure the same construct, such as four questions on the same attitude. Alpha near 1 means the items rise and fall together. Alpha near 0 means they do not.
Assumptions
Items are numeric and scored in the same direction. A reverse-worded item that was not reversed pulls alpha down and can make an item–total correlation negative. Rows with any missing item are dropped. Alpha is undefined if the total score does not vary.
Running it in Tensr
The menu key is reliability_cronbach. The alias reliability computes the same alpha and also flags an item whose correlation with the total is negative.
Options
Prop
Type
Reading the output
Five items built from one common factor plus unique noise. The common part is moderate, so alpha should land in the .70s rather than the .90s.
| Item | Mean | Std dev | Item–total r | α if deleted |
|---|---|---|---|---|
| item1 | -0.035 | 1.255 | 0.381 | 0.623 |
| item2 | -0.012 | 1.247 | 0.438 | 0.595 |
| item3 | -0.069 | 1.164 | 0.45 | 0.591 |
| item4 | 0.144 | 1.152 | 0.422 | 0.604 |
| item5 | -0.045 | 1.195 | 0.374 | 0.625 |
Cronbach's α for 5 items on 120 complete cases. Cronbach's α for 5 items on 120 complete cases. Metrics: Cronbach's α = 0.659; Items = 5; Cases = 120.
Reporting (APA 7)
Cronbach's α for 5 items on 120 complete cases. Report the estimate in the table. This procedure is not summarised by one p-value.
Coming from SPSS
Analyze → Scale → Reliability Analysis, model Alpha. SPSS also prints the item statistics. Tensr stores alpha, the item–total correlation, and alpha if that item is deleted. There is no split-half or Guttman model on this request.