HANDOUT 5.2:
MEASURES OF INTERNAL CONSISTENCY

TERM

SIGNIFICANCE
Cronbach's Alpha
A measure of internal consistency. Generally used for measures where participants respond to questions on a scale (1 to 3, 1 to 4, 1 to 5, etc.). Alpha can range between 0 and 1. If a scale has an alpha above .60, it is usually considered to be internally consistent.
Kuder-Richardson Reliability
Like Cronbach's alpha, except that it is used for measures in which items only can be responded to in one of two ways (true/false or yes/no). Like alpha, can range from 0 to 1. An internally consistent measure would have a Kuder-Richardson reliability of at least .60
Split-Half Reliability (often based on correlating the "Odds" with the "Evens")
An internally consistent measure should have a split half reliability of at least .75
Median Inter-Item Correlation
An index of the degree to which individual items correlate with each other. A median inter-item correlation of .30 usually indicates a high degree of internal consistency.
Factor Loadings
The degree to which an item appears to be correlated with a factor. Ideally, if test is supposed to measure two independent factors, half the items should load heavily (+.40-.80) on the first factor, but have near zero negative loadings on the second factor. The other should not load on the first factor, but load highly on second factor. Generally, you expect items measuring certain factor to have a loading of at least +.30 with factor.


Back to Chapter 5 Main Menu