COMM 2002 Lecture Notes - Lecture 6: Ordinal Data, Explained Variation
Document Summary
What is the direction of the relationship (+ or -) Strength of relationship between two nominal variables or one nominal and one ordinal variable. Coefficient between 0-1 no direction with nominal variables. Tells us whether knowing the value of one variable reduces the chance of making an error in predicting the value of another. Possible for all level of measurement but most commonly used on ordinal. Gamma, somers d, tau b and c. Nominal: chi-square and cramer"s v; logic frequency observed frequency of expected. Ordinal: gamma, somers d, tau b/c; pre logic reduction of error by knowing the independent variable; concordant pairs discordant pairs divided by con + dis (ties. The statistic used is pearson"s r often called the correlation coefficient. Pearson"s r is the strength measure of linear regression. R" tells us how closely correlated two interval level variables.