PSY201H1 Lecture Notes - Lecture 5: Homoscedasticity, Standard Deviation, Standard Error
Document Summary
Dividing what we can and can"t explain using rxy. For any given person: explainable + unexplainable = deviation score, (y" - )2 + (yi y")2 = (yi - )2 (y" - ) + (yi y") = (yi - ) Interpret like a standard deviation: e. g. , 68% of scores fall +/-1 se around the regression line. Less homoscedastic when distribution along regression line varies. Line that best fits scattered scores: smallest se possible when using x to predict y, maximizes ability to predict y using x, ensures only one line can fit. Least squares criterion: minimizes (y y")2, y" = predicted scores. Correlation: can"t manipulate variables (either for ethical or practical reasons, measure relationships among variables, no causation, e. g. Are high school grades related to success in university: e. g. Regression: can"t manipulate variables, measure relationships among variables, no causation, suspected causal direction. E. g. , temporal precedence: predictor (x values) and criterion (y values, more than one predictor, e. g.
