PSYC 2002 Lecture Notes - Lecture 13: Data Transformation, Phi Coefficient, Effect Size
9 Apr 2016
School
Department
Course
Professor
Document Summary
Regression: = bx + a, r2 (pearson r correlation coefficient, squared) or proportionate reduction in error. Error using mean as estimate of y (sstotal) vs. error using regression line as estimate of y (sserror) = [(o2/e)] n: chi-square ( ) test of independence, relationship be/w 2 variables (nominal) Effect size & power for nonparametic tests. Pros & cons of procedures for non normal distributions. Effect size can be determined for test of independence. Effect size conventions linked to 2 types of calculations done after statistic has been calculated: phi coefficient ( ): for 2 x 2 contingency tables, cramer"s phi: for larger than 2 x 2 tables. Power can be determined for test of independence. Power conventions can be used to determine approp sample size required given alpha, expected effect size, & # of categories being evaluated (df) Measurements made on interval or ratio scale.
