One-sample t-test output1 <- t.test(data,mu=0,alternative="two.sided") output2 <- t.test(data,mu=0,alternative="greater") output3 <- t.test(data,mu=0,alternative="less") ANDERSON-DARLING TEST FOR NORMALITY ad.test(data/data1/data2) LEVENE’S TEST FOR EQUAL VARIANCE data.sort <- stack(data[,-1]) *FULL DATA FILE, not data1 or data2 leveneTest(data.sort$values,data.sort$ind) boxplot(data1,data2) *equal if p-value>0.05 Binomial test for 1 proportion binom.test(x,n,p0,alternative="two.sided") binom.test(x,n,p0,alternative="greater") binom.test(x,n,p0,alternative="less") Regression coefficients → summary(model) Conf. interval → confint(model, level =0.95) Residual plot → plot(model) Scatter plot → model <- lm(y~x) → plot(x,y) Regression line → abline(model) One-way table data analysis chisq.test(data,p=prob) qchisq(1-alpha,n-1) Critical value sum(data)*prob expected value Two proportions z-test p1hat <- x1/n1 p2hat <- x2/n2 p.pool <- (x1+x2)/(n1+n2) z.star <- (p1hat-p2hat)/sqrt(p.pool*(1-p.pool)*(1/n1+1/n2)) pnorm(-abs(z.star)) * 2 Case 1 1 - pnorm(z.star) Case 2 pnorm(z.star) Case 3 mar.error <- qnorm(1-alpha/2)* sqrt(p1tilt*(1-p1tilt)/(n1+2) + p2tilt*(1p2tilt)/(n2+2)) stat - mar.error Lower bound stat + mar.error Upper bound Two-sample t-test # Case 1: mu1 =/ mu2 t.test(data1,data2,alternative="two.sided") # Case 2: mu1 > mu2 t.test(data1,data2,alternative="greater") # Case 3: mu1 < mu2 t.test(data1,data2,alternative="less") Matched-pair T-test*for normality, do DIFF (data1-data2) t.test(data1,data2,paired=TRUE,alternative="two.sided") t.test(data1,data2,paired=TRUE,alternative="greater") t.test(data1,data2,paired=TRUE,alternative="less") Mann-Whitney Test wilcox.test(data1,data2,exact=FALSE,alternative="two.sided") wilcox.test(data1,data2,exact=FALSE,alternative="greater") wilcox.test(data1,data2,exact=FALSE,alternative="less") Two-way data table analysis data.summary <- matrix(c(15,10,7,19,4,19),ncol=2,byrow=TRUE) rownames(data.summary) <- c("Desipramine","Lithium","Placebo") colnames(data.summary) <- c("No","Yes") chisq.test(data.summary) add “$observed” or “$expected” conditional distribution of Y given X → rowcond <scale(t(data.summary),center=FALSE,colSums(t(data.summary))) marginal distribution of Y → colSums(data.summary)/sum(data.summary) *** ADD “barplot” for bar chart conditional distribution of X → scale(t(data.summary),center=FALSE,colSums(t(data.summary))) conditional distribution of Y → scale(data.summary,center=FALSE,colSums(data.summary)) marginal distribution of X → rowSums(data.summary)/sum(data.summary) conditional distribution of X given Y → colcond <scale(data.summary,center=FALSE,colSums(data.summary)) CRITICAL VALUE for t-test # Case 1 -> H0: mu = mu0 and H1: mu =/ mu0 qt(alpha/2, df=n-1) qt(1alpha/2, df=n-1) # Case 2 -> H0: mu <= mu0 and H1: mu > mu0 qt(1-alpha, df=n-1) # Case 3 -> H0: mu >= mu0 and H1: mu < mu0 qt(alpha, df=n-1) Mann-Whitney Test-statistic dataagg <- c(data1,data2) # Case for n1 > n2 sum(rank(dataagg)[(n1+1):(n1+n2)]) # Case for n1 > n2 sum(rank(dataagg)[(n1+1):(n1+n2)]) # Case for n1 = n2 sum(rank(dataagg)[1:n1]) Large sample z-test for 1 proportion prop.test(x,n,p=p0,alternative="two.sided",correct=FALSE) prop.test(x,n,p=p0,alternative="greater",correct=FALSE) prop.test(x,n,p=p0,alternative="less",correct=FALSE) Wilcox signed test statistic # T.plus → wilcox.test(data1,data2,paired=TRUE,exact=FALSE, alternative="two.sided")$statistic # T.minus → n <- length(abs(diff)[diff!=0]) n*(n+1)/2 - wilcox.test(data1,data2,paired=TRUE,exact=FALSE, alternative="two.sided")$statistic NORMALITY CHECK hist(data), boxplot (data) qqnorm(data,ylim=c(min(data)-1,max(data)+1)) qqline(data,col="red") with(ToothGrowth, qqPlot(data, ylim=c(min(data)-1,max(data)+1), envelope = 0.95)) Wilcox signed test*for normality, do DIFF (data1-data2) wilcox.test(data1,data2,paired=TRUE,exact=FALSE,alternative="two.sided") wilcox.test(data1,data2,paired=TRUE,exact=FALSE,alternative="greater") wilcox.test(data1,data2,paired=TRUE,exact=FALSE,alternative="less")