PSY 302 Lecture Notes - Lecture 44: General Linear Model, Central Limit Theorem, Homoscedasticity
Document Summary
Data is randomly drawn if we had data in which null was true, and randomly drew a sample out of that population, how likely would we be to get an effect size from that. This assumption is pretty much never met. Nhst looks to see the probability of getting these results if the sample was randomly drawn. Large sample sizes (and central limit theorem) tend to ameliorate the negative effect of this can sidestep because of clt. Sometimes it is necessary to violate this . Sometimes necessary to use a non-parametric test (ex: spearmen"s r statistical analysis that don"t make assumptions about shape of data don"t assume that it is normal data: additional assumption of glm. Otherwise glm estimates tend to under estimate the strength of the relationship between variables. Sometimes a straight line doesn"t describe the data-sometimes need a curved line= different equations needed to do this.