PSYCH 100A Lecture Notes - Lecture 10: Null Hypothesis, Confidence Interval, Effect Size
Document Summary
More preferred in comparison to significance test. Observed sample stat - expected value under null hypothesis/ standard error of null hypothesis (comparison) distribution. Statistical tests: z-test, xbar - mu/ standard error (sigma xbar, t-test, xbar - mu/ standard error (s xbar, paired samples t-test, dbar - 0/ standard error (s dbar) Independent samples t-test a. t(df1+df2) = (xbar1 - xbar2) - 0/ standard error (s xbar1-xbar2) Make you are always using a two-tailed critical value. As confidence increases, say from 95% to 99%, confidence intervals become wider. Nhst evaluates the plausibility of a specific population value. Criticism: large sample sizes can affect the credibility of the significance test. Larger the sample sizes, p-value is farther from mu. All else being equal, as n goes up, standard error goes down, the t-test value goes up, and all effects eventually become statistically significant effects. Effect size will not change if n changes though.