NOTE: For your homework download and use the template (https://math.dartmouth.edu/~m50f17/HW2.Rmd)
Read the green comments in the rmd file to see where your answers should go.
Given a fixed confidence interval percentage (say 95%) at what value of x does CI on the mean response achieve its minimum width?
Write an R-chunk using the propellant data which computes the following.
The width of the interval is \[ 2 t_{\alpha/2,n-2} \sqrt{MS_{Res} ((1/n) + (x_0 - \bar{x})^2 / S_{xx} } \] and all terms inside the square root are positive. Therefore it is minimized when \(x_0=\bar{x}\).
# Computation part of the answer :
prop<-read.table("https://math.dartmouth.edu/~m50f17/propellant.csv", header=T, sep=",")
shearS<-prop$ShearS
age<-prop$Age
plot(age, shearS, xlab = "Propellant Age (weeks)", ylab = "Shear S. (psi)", main = "Rocket Propellant")
fitted <- lm(shearS ~ age)
ageList <- seq(0,25,0.5)
cList <- predict(fitted, list(age = ageList), int = "c", level = 0.95)
pList <- predict(fitted, list(age = ageList), int = "p", level = 0.95)
matlines(ageList, pList, lty='solid' , col = "red")
matlines(ageList, cList, lty = 'solid', col = "blue")
# since n=20 we look at the t_18 distribution
wantedQuantile <- qt( 0.95, 18) ;
cat("95% quantile is of t_18 is : ", wantedQuantile ) ;
## 95% quantile is of t_18 is : 1.734064
Plot the same graph as in Question-1 without using R function predict, but instead directly calculating the interval limits we discussed in class. In particular, what are the limits of 95% confidence interval on \(E(y|x_0)\)?
# Computation part of the answer :
Load the propellant data and fit a simple linear regression model relating shear strength to age.
# Computation part of the answer :
Load the propellant data. This time let us consider a relation between square of shear strength and propellant age.
# Computation part of the answer :
Once again using propellant data fit a simple linear regression model between shear strength and propellant age. Consider the steps that we used to obtain t-test for hypothesis \(\beta_1 = G_1\), and following similar steps in order to develop a test for \(\beta_1 > G_1\) instead. Then
# Computation part of the answer :