Reliability, Cronbach’s Alpha

Reliability and Measurement of Reliability

Reliability in general relates to how much error of measurement there is in a particular measuring process. Error of measurement relates to random differences that occur in the process that are not related to the variable being measured. These differences are known as random measurement error because they differ from person to person and cannot be easily predicted. The more random measurement error a measure of a variable has, the lower the reliability of that measurement. Examples of random measurement errors in scales might include writing double-barreled questions (asking more than one question in a “single” question), including words in questions that some participants do not understand, or providing unclear directions for a scale.

There are several ways to measure reliability, including test–retest reliability (having individuals answer a scale at different times and examining whether their answers are the same across all items), parallel forms reliability (having individuals answer two versions of a scale at different times and examining how well the two versions agree), and intercoder reliability (whereby researchers have trained individuals to observe or code human interactions or text related to the variables being measured and examining how well the coders agree). This entry discusses the internal consistency measure of reliability, which involves using a scale of multiple items to measure the same variable at one point in time. Because all the indicators of the variable are measured at once, reliability for these variables relates to the consistency of the measure’s items across the scale rather than the stability of the measure across time (which a more longitudinal design, such as test–retest reliability or parallel forms reliability would measure).

...