Figure 3.
Test-retest reliability effects on observed group differences. The top row panels (A–C) illustrate the different analysis scenarios, while the 2 row panels (D–F) show the corresponding observed effects for different error and between-subject variance values of task performance, and the bottom row panels (G–I) show the corresponding observed effects for different error and between-subject variance values of symptoms/traits. (A) An illustration of correlation between traits or symptoms and task performance. The vertical dashed lines indicate how the data was split for the two analysis scenarios. (B) An illustration of patient and control groups created by assigning 10% of the population with the poorest mental health to patient group and the remaining 90% to control group. (C) An illustration of “low” and “high” trait groups by performing a median split. Overall, the test results show that both observed correlation strength and observed group differences increase with increasing test-retest reliability (i.e., with reducing error variance/increasing between-subject variance) both when varying between-subject and error variances of task performance measures (D–F) and symptoms/traits measures (G–I).
