PS295 Lecture Notes - Lecture 6: Phi Coefficient, Level Of Measurement, Contingency Table
Document Summary
Perfect positive relation: scores go up by equal amount towards the right. The coefficient of determination (r-squared: correlation coefficient squared, proportion of variance in one variable that is accounted for by the other variable, multiplying by 100 gives percentage, often called variance explained . Regression: regression analysis is a technique for finding the line of best fit through a data set, the regression line minimizes the squared distance of points from the line. Factors the distort correlation coefficients: restricted range, outliers, reliability of measures. Restricted range: when sample under study does not include the full range of the variables, sample is relatively homogenous for one of the variables, the apparent association between the variables can be greatly reduced. Outliers: an extreme or highly unusual score; a score that is deviant from the remainder of the data, can have big impact on correlation coefficient, off-line outliers, on-line outliers. Reliability of measures: unreliable measures attenuate (reduce) the correlation coefficient.