t7q1 <- read.table("C:/Users/Wei Hao/Desktop/ST2137/Tutorials/Data/locate.txt", header=T)
attach(t7q1)
model1 <- aov(sales~location)
group.1 <- sales[location=="F"]
group.2 <- sales[location=="M"]
group.3 <- sales[location=="R"]
group.means <- tapply(sales,location,mean)
treat.group <- cbind(group.1,group.2,group.3)
mse <- sum(model1$res^2)/15
lsd <- qt(0.975,15)*sqrt(mse*2/6)
check.lsd<-function(obj,i,j,lsd){
mx <- mean(obj[,i])
my <- mean(obj[,j])
d <- mx - my
if(abs(d)>lsd)
cat("There is siginificant difference between groups",i,"&",j, "\n",
"Means =",mx,",",my," Diff =", d," > LSD =",lsd,"\n")
else
cat("There is no siginificant difference between groups",i,"&",j,"\n",
"Means =",mx,",",my," Diff =", d," < LSD =",lsd,"\n")
}
Let \(\mu_1, \mu_2m, \mu_3\) be the average sales for stores with front, middle and rear aisle locations respectively. We are interested to test \(H_0: \mu_1 = \mu_2 = \mu_3\) against \(H_1: \mu_i \neq \mu_j\), for some \(i \neq j\).
check.lsd(treat.group,1,2,lsd)
## There is siginificant difference between groups 1 & 2
## Means = 6.066667 , 2.266667 Diff = 3.8 > LSD = 1.600086
check.lsd(treat.group,1,3,lsd)
## There is siginificant difference between groups 1 & 3
## Means = 6.066667 , 3.733333 Diff = 2.333333 > LSD = 1.600086
check.lsd(treat.group,2,3,lsd)
## There is no siginificant difference between groups 2 & 3
## Means = 2.266667 , 3.733333 Diff = -1.466667 < LSD = 1.600086
contrasts(location) <- matrix(c(0,-1,1,2,-1,-1), nrow=3)
contrasts(location)
## [,1] [,2]
## F 0 2
## M -1 -1
## R 1 -1
modelc<-aov(sales~location)
summary.lm(modelc)
##
## Call:
## aov(formula = sales ~ location)
##
## Residuals:
## Min 1Q Median 3Q Max
## -2.06667 -0.86667 -0.06667 0.73333 2.53333
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 4.0222 0.3065 13.124 1.26e-09 ***
## location1 0.7333 0.3754 1.954 0.069643 .
## location2 1.0222 0.2167 4.717 0.000275 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 1.3 on 15 degrees of freedom
## Multiple R-squared: 0.6347, Adjusted R-squared: 0.586
## F-statistic: 13.03 on 2 and 15 DF, p-value: 0.0005242
Test statistic \(F = 13.03\), p-value \(= 0.0005\). Since \(F_{obs} = 13.03\) is above the critical bound of \(F = 3.68\), reject \(H_0\). Alternatively, the p-value is smaller than \(0.05\), so we reject \(H_0\). There is enough evidence to conclude that the average sales volumes in thousands of dollars are different across the three store aisle locations.
We see that \(|\bar{x_1 } - \bar{x_2 } | > LSD\) and \(|\bar{x_1 } - \bar{x_3 } | > LSD\) so the front aisle location is different from the other two locations while the middle and rear locations are not different in terms of the average sales volumes.
See SAS code outputs and solutions.