Health Sciences 2801A/B Lecture Notes - Lecture 6: Observational Error, Confidence Interval, Emor
Document Summary
Useful to compare a subsequent score obtained after treatment with initial score obtained before treatment. If treatment has beneficial effect, we expect better score after treatment. However, important to recognize that both scores are affected by measurement error: as a result, subsequent and initial score are expected to differ, even when treatment has no effect. Confidence interval is practical method of addressing effect of measurement error in situation where person is tested twice. Two types of measurement error: unsystematic error - type of measurement error that increases/decreases individual scores by unpredictable amount. Score is equally likely to increase/decrease as a result of unsystematic error. The effect of positive systematic bias to move upper bound of confidence interval in upward direction. Negative systematic bias is to move lower bound of confidence interval in downward direction: thus, combined effect of systematic bias and unsystematic error is addressed by confidence interval that extends unequally above and below a score.