PSY309H5 Lecture Notes - Lecture 10: Randomized Experiment, Analysis Of Variance, Internal Validity
Document Summary
One-sample t-test, chi-square tests: descriptive > summarize data. Simple experiments (t-tests), one independent variable, more than three levels (anova), factorials (anova): randomized experiment > determine causes. Basic considerations: power: the probability that your test correctly rejects the null hypothesis when the null is false. ): reduces your standard error, or the denominator in your tests. Reduce your sample error variance: have a tightly controlled experiment with reliable measures. Have a directional hypothesis: one-tailed vs. two-tailed test. Effect size: manipulations that are more extreme from one another - a louder signal (also increases the numerator in your test) Select a more homogenous population: reduces the population standard error. Change alpha from . 05 to . 10: but . Scales of measurement: important implications for analysis. Validity of your study: internal validity - history, selection bias, instrumentation, testing, external validity - generalization, artifacts or bias - subject: demand characteristics, reactivity; experimenter: expectancy effects, statistical conclusions - power, test assumptions, error state.
