#1.) library(Rfit) Y=telephone$calls x=telephone$year plot(Y~x,xlab='Rok',ylab='Pocty hovoru') #(a) fit<-rfit(Y~x) summary(fit) #(b) summary(lm(Y~x)) plot(Y~x,xlab='Rok',ylab='Pocty hovoru') abline(lm(Y~x)$coef) abline(fit$coef,col=2) #(c) n=length(x) Sn=function(b){ R=rank(Y-x*b) Sn=sum((x-mean(x))*(R/(n+1)-1/2)) return(Sn) } beta_hat=uniroot(Sn,c(-1,1))$root beta_hat #(d) Dn=function(b){ R=rank(Y-x*b) Dn=sum((Y-(x-mean(x))*b)*(R/(n+1)-1/2)) } beta_hat2=optimize(Dn,c(-1,1))$min beta_hat2 #(e) e=Y-beta_hat*x signedrank(e) ######################## #(2) library(quantreg) #(a) Y=bbsalaries$logSalary x1=bbsalaries$aveWins x2=bbsalaries$aveGames n=length(Y) fit<-rfit(Y~x1+x2) summary(fit) #(b) Tn=function(b){ R=rank(Y-x1*b[1]-x2*b[2]) Sn1=sum((x1-mean(x1))*(R/(n+1)-1/2)) Sn2=sum((x2-mean(x2))*(R/(n+1)-1/2)) Tn=sqrt(Sn1^2+Sn2^2) #Tn=abs(Sn1)+abs(Sn2) #Tn=max(abs(Sn1),abs(Sn2)) } optim(c(0,0),Tn) #(c) Dn=function(b){ R=rank(Y-x1*b[1]-x2*b[2]) Dn=sum((Y-(x1-mean(x1))*b[1]-(x2-mean(x2))*b[2])*(R/(n+1)-1/2)) } optim(c(0,0),Dn) #rrs.test(x1,x2,Y) #(d) H0: beta1=0 a=rq(Y~x2, tau=-1) b=ranks(a)$ranks X=cbind(rep(1,n),x2) Z_hat=X%*%solve(t(X)%*%X)%*%t(X)%*%x1 Dn_hat=sum((x1-Z_hat)^2)/(n-1) Tn=sum((x1-Z_hat)*b)/sqrt(n-1)/sqrt(Dn_hat)*sqrt(12) 2*(1-pnorm(Tn)) # aligned test H0: beta1=0 Dn=function(b){ R=rank(Y-(x2-mean(x2))*b) Dn=sum((Y-(x2-mean(x2))*b)*(R/(n+1)-1/2)) } beta2_hat=optimize(Dn,c(0,10))$min e=Y-x2*beta2_hat R_hat=rank(e) Dn=var(x1) T1=sum((x1-mean(x1))*R_hat/(n+1))/sqrt(n-1)/sqrt(Dn)*sqrt(12) 2*(1-pnorm(abs(T1))) #(e) H0: beta2=0 a=rq(Y~x1, tau=-1) b=ranks(a)$ranks X=cbind(rep(1,n),x1) Z_hat=X%*%solve(t(X)%*%X)%*%t(X)%*%x2 Dn_hat=sum((x2-Z_hat)^2)/(n-1) Tn=sum((x2-Dn_hat)*b)/sqrt(n-1)/sqrt(Dn_hat)*sqrt(12) 2*(1-pnorm(abs(Tn))) #aligned test H0: beta2=0 Dn=function(b){ R=rank(Y-(x1-mean(x1))*b) Dn=sum((Y-(x1-mean(x1))*b)*(R/(n+1)-1/2)) } beta1_hat=optimize(Dn,c(0,10))$min e=Y-x1*beta1_hat R_hat=rank(e) Dn=var(x2) Tn=sum((x2-mean(x2))*R_hat/(n+1))/sqrt(n-1)/sqrt(Dn)*sqrt(12) 2*(1-pnorm(abs(Tn)))