PSYC 3130 Lecture Notes - Lecture 9: Null Hypothesis
Document Summary
Biggest source of error strains from individual differences. This is why related samples t- tests are useful (decreases the denominator of the f statistic, which increases power) Unfortunately, there are carry-over and order effects in related samples t-tests. Homogeneity of variance is a nonissue in related samples tests. Assumed that groups are non-random/dependent (scores can be predicted in the other group) Assumed that group variances are likely correlated. One group may be measured twice or two groups may be paired together. Under the null hypothesis, the assumption is that there is zero difference between correlated means. Under the alternative, we conclude that the true difference is at least non-zero. We may have one variable called a difference variable. Two critical values in the tails that are unlikely to occur if null is true. N-1 is used since the only group is the difference variable group.