PSY309H5 Lecture Notes - Lecture 3: Convenience Sampling, Internal Consistency, Longitudinal Study
Document Summary
Psy309: lecture 3: low reliability means there is no validity. Rule of thumb is that reliability needs to be more than 0. 70. Actual correlation is 0. 40, our simulation can only go up to 0. 39 which is still very good: example 2: non-reliable measures (1=0. 80 and 2=0. 30) Actual correlation is 0. 4, our simulation can only give us a correlation up to 0. 20 which is not a good correlation as our test is probably only measuring noise. The noise is attenuating (reducing) the reliability. Assessing reliability: test-retest reliability; consistent over time. If the same subject/participant is tested at one point and again at a later point, if the results are similar, test-retest reliability is high. You take score-time 1 and predict score-time 2 using this score. If score-time 2 is similar to prediction, test-retest reliability is high: when not to use test-retest reliability. When attributes are not stable, test-retest reliability should not be used eg: mood.