vote=matrix(c(794,86,150,570),nr=2,dimnames=list("1st Survey"=c("Approve","Disapprove"),"2nd Survey"=c("Approve","Disapprove"))) vote ##Set correct=F to apply McNemar without continuity correction mcnemar.test(vote,correct=F) ##Set correct=t to apply McNemar with continuity correction ##Note: we don't really need the correction here since the sample size is large mcnemar.test(vote,correct=T) ### Simple Kappa Coefficient ### Using the original formula to calculate the Simple Kappa Coefficient prop=vote/sum(vote) Po=sum(diag(prop)) Pe=rowSums(prop)[1]*colSums(prop)[1]+rowSums(prop)[2]*colSums(prop)[2] kappa=(Po-Pe)/(1-Pe) kappa ### Using a function Kappa() in package vcd. ### Please first load packages VR, colorspace and grid and finally load vcd. #install.packages("vcd") library(vcd) kappa=Kappa(vote) CI_kappa=cbind(0.69959267-qnorm(0.975)*0.01805518,0.69959267+qnorm(0.975)*0.01805518) CI_kappa ### or use confint() function confint(kappa) ###agreement plot ####observed and expected diagonal elements are represented by superposed black and white rectangles, respectively. agreementplot(vote) ####Consider cross-sectional design vote=matrix(c(944,880,656,720),nr=2,dimnames=list(c("1st Survey","2nd Survey"),c("Approve","Disapprove"))) vote count=vote ##The ususal chi-sq. test result=chisq.test(vote, correct=F) result ##Set correct=F to apply McNemar without continuity correction mcnemar.test(vote,correct=F) ##Set correct=t to apply McNemar with continuity correction ##Note: we don't really need the correction here since the sample size is large mcnemar.test(vote,correct=T) Contingency_Table=list(Frequency=count,Expected=result\$expected,Deviation=count-result\$expected,Percentage=prop.table(count),RowPercentage=prop.table(count,1),ColPercentage=prop.table(count,2)) Contingency_Table