PSY 370 Lecture Notes - Lecture 32: Internal Consistency, Psy, Factor Analysis
Document Summary
Each type of reliability assesses a different source of error. Need to report what kind(s) of reliability you tested. This means that 70% of the variance is true score variance. Most people will accept a reliability coefficient of above . 70. For making decisions about people (and peoples" lives) Other people argue that this indicates redundancy among the items. The higher the stakes, the higher your reliability should be. More items = higher reliability (unless the items really don"t fit) But you can get a high reliability coefficient with a heterogeneous test. Having a high reliability coefficient does not necessarily mean your items are all measuring the same thing. Need factor analysis or other methods to test homogeneity. Text calls this static vs. dynamic not quite so simple. If you have very little variance in responses, your reliability coefficient will be: Reliability analysis is based on correlations, which are based on variances.
