Skip to main content
Glite Dictionary
split-half correlation

split-half correlation

9.4
A number that shows how similar scores are on two parts of a test
  • noun
  • /splɪt hɑːf ˌkɔːrəˈleɪʃən/
  • Specialized
translation icon : correlación de mitades divididas
  • A determination of the reliability of the instrument was made by performing a split-half correlation analysis on the data obtained in this study.

Examples

  • The reliability of the Rod-and-Frame Test was checked with split-half correlation, with a coefficient of .73 (p < .001).

    Academic text (2008)

  • The data showed that the alpha for the total WGCTA-FS was .76 and the split-half correlation was r = .44.

    Academic text (2005)

  • Explore produced a split-half correlation of .60, and all remaining correlations were weak.

    Academic text (2006)

  • The researchers calculated the split-half correlation to evaluate the consistency of test scores from two different groups of participants.

  • To assess the test's reliability, a split-half correlation was performed, resulting in a strong coefficient of .85.

  • By analyzing the split-half correlation, the psychologists were able to confirm the stability of the measurement tool over time.

Synonyms

chance-half correlation
vssplit-half correlation
  • Specialized
9.8

A number that shows how similar scores are between two random halves of a test

uses two halves that may not be randomly chosen

Surface Forms

split-half correlation singular
split-half correlations plural

Morphology

split-half + correlation

The term literally combines 'split' + 'half' (divide into two parts) with 'correlation' (a statistical relationship), so a learner who knows those words can infer it denotes the correlation between two halves of a test. Although domain-specific, the composition is direct and predictable rather than idiomatic.

Etymology

Split-half correlation comes from the simple idea of making a split of a test into two halves and then checking how closely the scores from each half match. If the two halves give similar results, the test is seen as 'consistent', so the measure tells us how 'reliable' the test is today.