Use the data set below Choose an observation index i eg i 3
Use the data set below. Choose an observation index i (e.g. i = 33, which corresponds to the outlying observation number 33) and create an indicator (dummy) variable Ui, where all the values of Ui are zero except for its ith value which is one. Now consider comparing the following models:
H0 : Time = B0 + B1Distance + B2Climb + E (Model 4.27)
H1 : Time = B0 + B1Distance + B2Climb + B3Ui + E (Model 4.28)
Let r*i be the ith externally standardized residual obtained from fitting Model 4.27. Show (or verify using an example) that:
The t-test for testing B3 = 0 in Model 4.28 is the same as the ith externally standardized residual obtained from Model 4.27, that is, t3 = r*i.
The F-test for testing Model 4.27 versus 4.28 reduces to the square of the ith externally standardized residual, that is, F = r*i^2.
Fit Model 4.27 to the data set without the ith observation.
Show that the estimates of B0, B1, and B2 in Model 4.28 are the same as those obtained in c. Hence adding an indicator variable for the ith observation is equivalent to deleting the corresponding observation.
Hill Race Time Distance Climb
Greenmantle New Year Dash 965 2.5 650
Carnethy 2901 6 2500
Craig Dunain 2019 6 900
Ben Rha 2736 7.5 800
Ben Lomond 3736 8 3070
Goatfell 4393 8 2866
Bens of Jura 12277 16 7500
Cairnpapple 2182 6 800
Scolty 1785 5 800
Traprain Law 2385 6 650
Lairig Ghru 11560 28 2100
Dollar 2583 5 2000
Lomonds of Fife 3900 9.5 2200
Cairn Table 2648 6 500
Eildon Two 1616 4.5 1500
Cairngorm 4335 10 3000
Seven Hills of Edinburgh 5905 14 2200
Knock Hill 4719 3 350
Black Hill 1045 4.5 1000
Creag Beag 1954 5.5 600
Kildoon 957 3 300
Meall Ant-Suiche 1674 3.5 1500
Half Ben Nevis 2859 6 2200
Cow Hill 1076 2 900
North Berwick Law 1121 3 600
Creag Dubh 1573 4 2000
Burnswark 2066 6 800
Largo 1714 5 950
Criffel 3030 6.5 1750
Achmony 1257 5 500
Ben Nevis 5135 10 4400
Knockfarrel 1943 6 600
Two Breweries Fell 10215 18 5200
Cockleroi 1686 4.5 850
Moffat Chase 9590 20 5000
Solution
