PSY 2106 Lecture Notes - Lecture 18: Chi-Squared Distribution, Psychokinesis, Count Data
Document Summary
100 people flipping a coin 20 times. No longer have 100 people flipping a coin 20 times. Instead we have two seperate groups of 20 people, one control group and one psychokinesis group, with each person flipping a coin once. Critical difference: we"re now using categorical data. Logically, the analysis is the same: we look at the theoretical distribution of the statistic to determine the probability of a difference using frequencies. Not based on parameters of the normal distribution (means, standard deviation) Many non-parametric stats make no assumptions about distribution- rather, based purely on probability. If you can, collect data using a continuous variable, do. Parametric tests (eg anova, t-tests) have greater power and continuous variables are more reliable than categorical. Still, there are many situation in which categorical data is the best available: Use tables, taking df and alpha intro account. We will use the x2 test statistic for. X2 test of independence (2 or more factors)