Transcript 10/20/2015
Psychology 202a Advanced Psychological Statistics October 20, 2015 Effect sizes • Not necessarily standardized. • Evil journal editors. • Standardized effect sizes corresponding to one-sample tests. d = (M – m0 ) / s • Effect sizes for differences between means. d = (M1 – M2) / sP • What’s large? What’s small? It depends. Confidence intervals • Goal: to find a range of reasonable values for a parameter based on a statistic and its sampling distribution. • Question: what null hypotheses would I not reject on the basis of these data? • The 100 * (1-a) % CI for the mean: M ± t critical × s M Things the 95% confidence interval doesn't mean • There is a 95% probability that m is between (lower bound) and (upper bound). • I am 95% certain (or confident) that m is between (lower bound) and (upper bound). • There is a 95% chance that the interval (lower bound) to (upper bound) contains m. • If I conduct confidence intervals in this way, 95% of the time, m will be between (lower bound) and (upper bound) Things the 95% confidence interval does mean • There is a 95% chance that the confidence interval resulting from the sample I am about to take will contain m. • If I conduct intervals in this way, 95% of the time, whatever interval I get will contain m. • Therefore, it is reasonable to act as if this interval contains m... • ...because if I do so, in the long run, I will act correctly 95% of the time. • Confidence intervals and duck-hunting statisticians.