A comparison of the independent-samples t-test and the paired-samples t-test when the observations are nonnegatively correlated pairs

Moshe Pollak*, Jiska Cohen

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

Abstract

An account of the behavior of the independent-samples t-test when applied to homoschedastic bivariate normal data is presented, and a comparison is made with the paired-samples t-test. Since the significance level is not violated when applying the independent-samples t-test to data which consist of positively correlated pairs and since the estimate of the variance is based on a larger number of 'degrees of freedom', the results suggest that when the sample size is small, one should not worry much about the possible existence of weak positive correlation. One may do better, powerwise, to ignore such correlation and use the independent-samples t-test, as though the samples were independent.

Original languageEnglish
Pages (from-to)133-146
Number of pages14
JournalJournal of Statistical Planning and Inference
Volume5
Issue number2
DOIs
StatePublished - 1981

Keywords

  • Power
  • Size
  • t-test

Fingerprint

Dive into the research topics of 'A comparison of the independent-samples t-test and the paired-samples t-test when the observations are nonnegatively correlated pairs'. Together they form a unique fingerprint.

Cite this