Study over all material since the last exam, including the material on R functions that we have covered. A lot of this material is on confidence intervals and tests of hypotheses. On tis cheat sheet I have summarized this material in a simple table form. Estimates & Confidence Intervals Recall that a (1−α)100% confidence interval for a paremter traps the true parameter (1−α)100% of the time. Table 0.1 displays the estimates and confidence intervals that we have considered so far. Note that the R-code for each is given. The CI and estimate can be saved. For example, f it = t.test(x), est = f it$est and vecci = f it$conf.int. Table 0.1: Table of Estimates and Confidence Intervals (so far). Note that IS stands for independent samples and PS stands for paired samples. µ Sample(s) Estimate CI R-code x= c(x1 , . . . , xn ) x x ± tα/2,n−1 √sn t.test(x) Based on WA’s q p̂ ± zα/2 p̂q̂ n wilcox.test(x,conf.int=TRUE) x−y x − y ± tα/2,n1 +n2 −2 SE q 2 s s2 SE = n11 + n22 t.test(x,y) med{Xi − Yj } Based on diffs. wilcox.test(x,y,conf.int=TRUE) d d ± tα/2,n−1 √sdn t.test(d) Based on WA’s wilcox.test(d,conf.int=TRUE) p̂1 − p̂2 ± zα/2 SE q SE = p̂n1 q̂11 + p̂n2 q̂22 prop.test(c(S1,S2),c(n1,n2)) p̂1 − p̂2 ± zα/2 SE q 2 1 −p̂2 ) SE = p̂1 +p̂2 −(p̂ n Class function µ x= c(x1 , . . . , xn ) medi≤j p S = #Suc p̂ = n xi +xj 2 o S n prop.test(S,n) n = samp.size µ 1 − µ2 x= c(x1 , . . . , xn1 ) IS y= c(y1 , . . . , yn2 ) µ 1 − µ2 x= c(x1 , . . . , xn1 ) IS y= c(y1 , . . . , yn2 ) µ 1 − µ2 di = xi − yi PS d= c(d1 , . . . , dn ) µ 1 − µ2 di = xi − yi PS d= c(d1 , . . . , dn ) p1 − p2 Succ. S1 & S2 p̂1 = S1 n1 IS Based on n1 & n2 p̂2 = S1 n2 p1 − p2 Succ. S1 & S2 p̂1 = S1 n PS Based on n p̂2 = S1 n medi≤j n di +dj 2 o 1 Tests • Level of test = PH0 [Reject H0 ]. • Power of test at θ = Pθ [Reject H0 ]. • p-value of test is the lowest level, based on the data, at which the test rejects H0 . Tables 0.2 and 0.3 displays the confidence intervals that we have considered so far. Note that the R-code for each is given. 2 Table 0.2: Table of Tests (so far). Note that IS stands for independent samples and PS stands for paired samples. µ Sample(s) Test Stat x= c(x1 , . . . , xn ) t= x−µ √0 s/ n Null Alt Reject at α H0 : µ = µ0 HA : µ > µ0 pvalue ≤ α HA : µ < µ0 pvalue ≤ α HA : µ 6= µ0 pvalue ≤ α Rcode: tst = t.test(x, mu = mu0, alternative = c(”greater”, ”less”, ”two.sided”)) P µ x= c(x1 , . . . , xn ) T = xi −µ0 >0 R(|xi |) H0 : µ = µ0 HA : µ > µ0 tst$p.value pvalue ≤ α HA : µ < µ0 pvalue ≤ α HA : µ 6= µ0 pvalue ≤ α Rcode: tst = wilcox.test(x, mu = mu0, alternative = c(”greater”, ”less”, ”two.sided”)) tst$p.value p HA : p > p0 pvalue ≤ α HA : p < p0 pvalue ≤ α HA : p 6= p0 pvalue ≤ α S = #Suc S H0 : p = p0 n = samp.size Rcode: tst = prop.test(S, n, p = p0, alternative = c(”greater”, ”less”, ”two.sided”)) µ1 − µ2 t = x−y SE q 2 s SE = n11 + tst$p.value HA : µ1 > µ2 pvalue ≤ α HA : µ1 < µ2 pvalue ≤ α HA : µ1 6= µ2 pvalue ≤ α Rcode: tst = t.test(x, y, alternative = c(”greater”, ”less”, ”two.sided”)) P 1 µ1 − µ2 x= c(x1 , . . . , xn1 ) W = ni=1 R(xi ) H0 : µ1 = µ2 HA : µ1 > µ2 tst$p.value IS HA : µ1 < µ2 pvalue ≤ α HA : µ1 6= µ2 pvalue ≤ α IS x= c(x1 , . . . , xn1 ) y= c(y1 , . . . , yn2 ) H0 : µ1 = µ2 s22 n2 y= c(y1 , . . . , yn2 ) Rcode: tst = wilcox.test(x, y, alternative = c(”greater”, ”less”, ”two.sided”)) 3 pvalue ≤ α tst$p.value Table 0.3: Table of Tests (so far). Note that IS stands for independent samples and PS stands for paired samples. Sample(s) Test Stat Null Alt Reject at α µ1 − µ 2 di = xi − yi t= H0 : µ1 = µ2 HA : µ1 > µ2 pvalue ≤ α PS d= c(d1 , . . . , dn ) HA : µ1 < µ2 pvalue ≤ α HA : µ1 6= µ2 pvalue ≤ α Rcode: tst = t.test(d, alternative = c(”greater”, ”less”, ”two.sided”)) P µ1 − µ2 di = xi − yi T = di >0 R(|di |) H0 : µ1 = µ2 HA : µ1 > µ2 tst$p.value PS HA : µ1 < µ2 pvalue ≤ α HA : µ1 6= µ2 pvalue ≤ α d √ s/ n d= c(d1 , . . . , dn ) Rcode: tst = wilcox.test(d, alternative = c(”greater”, ”less”, ”two.sided”)) p1 − p2 Succ. S1 & S2 z= √ pq ˆ1 −p̂2 p̂q̂ n1 + n1 1 IS p̂ = H0 : p1 = p2 HA : p1 > p2 pvalue ≤ α n1 p1 +n2 p2 n1 +n2 HA : p1 < p2 pvalue ≤ α HA : p1 6= p2 pvalue ≤ α Rcode: tst = prop.test(c(S1, S2), c(n1, n2), alternative = c(”greater”, ”less”, ”two.sided”)) Succ. S1 & S2 PS Based on n −pˆ2 z = p̂1SE q 2 SE = p̂1 +p̂ n tst$p.value 2 q̂ = 1 − p̂ p1 − p2 pvalue ≤ α H0 : p1 = p2 Rcode: tst = Student Function tst$p.value HA : p1 > p2 pvalue ≤ α HA : p1 < p2 pvalue ≤ α HA : p1 6= p2 pvalue ≤ α tst$p.value 4