H0: 1 2 H0: 1 = 2 H0: 1 2 H1: 1 < 2 H1: 1 2 H1: 1 > 2 p1 p2 x1 x2 combined number of successes Test statistic: z where pc 1 1 n 1 n2 combined sample size pc (1 pc ) n1 n2 Normality may be assumed if each sample has at least 10 “successes” and 10 “failures,” i.e., if n1p1 10, n1(1p1) 10, n2p2 10, n2(1p2) 10 (some textbooks say 5 instead of 10).
Test for Equality of Two Means (Independent Samples)
Left-Tail Test Two-Tail Test Right-Tail Test H0: 1 2 H0: 1 = 2 H0: 1 2 H1: 1 < 2 H1: 1 2 H1: 1 > 2 Case 1: Known Variances or Very Large Samples (z-test) x1 x2 x1 x2 zcalc or zcalc Use z-table for critical values. 1 2 2 2 2 s1 s2 2 n1 n2 n1 n2
Case 2: Unknown Variances Assumed Equal (pooled samples t-test)
x1 x2 (n 1) s12 (n2 1) s2 2 tcalc with pooled variance s p 2 1 1 1 n1 n2 2 s p2 n1 n2 Use d.f. = (n11)+(n21). If both samples are large (n1 30 and n2 30) see Situation 1. Case 3: Unknown Variances Assumed Unequal (Welch’s t-test) 2 x1 x2 s12 n1 s2 2 n2 tcalc Use d.f. s s 2 2 s12s 2 2 n1 2 n2 2 1 2 n1 n2 n1 1 n2 1 Welch’s d.f. formula is too hard for hand calculations, but MINITAB and Excel use it. The test statistics in Case 2 and Case 3 will be the same if n1 = n2. (but decisions may differ due to d.f.).
Test for Equality of Two Variances
Left-Tail Test Two-Tail Test Right-Tail Test H0: 1 2 2 2 H0: 12 = 22 H0: 12 22 H1: 12 < 22 H1: 12 22 H1: 12 > 22 or, equivalently, H0: 12/22 1 H0: 12/22 = 1 H0: 12/22 1 H1: 1 /2 < 1 2 2 H1: 12/22 H1: 12/22 > 1 Test statistic is Fcalc s12 s22 with df1 = n11 (numerator) and df2 = n21 (denominator). The F test is sensitive to population skewness. If you put the larger variance in the numerator (to force a right-tail test) then use when you look up the critical value in the F table.