PSYC 301 Lecture Notes - Lecture 21: Type I And Type Ii Errors, Null Hypothesis, Sphericity
Document Summary
If you look at difference scores between any two columns in the design, the population variance of that difference score will be constant regardless of which two columns you"re comparing. Offi(cid:272)ial assu(cid:373)ptio(cid:374) is (cid:272)o(cid:373)pou(cid:374)d sy(cid:373)(cid:373)etry (cid:272)o(cid:448)aria(cid:374)(cid:272)e (cid:271)et(cid:449)ee(cid:374) a(cid:374)y (cid:272)olu(cid:373)(cid:374)s of s(cid:272)ores is equal. Almost impossible to satisfy, so as long as the simple assumption is satisfied, that"s good enough, and the test will be accurate. Book 1 book 2 book 3 row mean. Column mean y. 1 = 5. 8 y. 2 = 7. 6 y. 3 = 7. 2 y = 6. 867. Sphericity assumes that error is constant across all of the scores. Sphericity assumption is realistically never met in practice. When the sphericity assumption is not met, the f-statistic will be too liberal. Multiply both numerator and denominator degrees of freedom by. Results in df numerator equal to 1 and df denominator equal to (n-1) This method is overly conservative but easy to implement.
