In: Statistics and Probability
The accompanying data table shows the energy consumed (in millions of Btu) in one year for a random sample of households from four regions. At alpha equals 0.10, can you conclude that the mean energy consumption of at least one region is different from the others? Perform a one-way ANOVA test by completing parts a through d. If convenient, use technology to solve the problem. Assume that each sample is normally distributed and that the population variances are equal. LOADING... Click to view the data table of regional energy consumption.
North West Midwest South West
136.1 132.1 56.5 124.8
109.5 79.3 63.2 58.3
101.5 61.6 50.6 46.2
114.6 98.6 51.3 108.4
160.3 68.8 127.7
182.0
Null hypotheses H0 : The average energy consumption is same for all regions.
Alternative hypotheses H1 : At least one average energy consumption is different among the four regions.
Level of significance = 0.10
Degree of freedom of group = Number of level - 1 = 4 - 1 = 3
Degree of freedom of error = Number of observations - Number of level = 20 - 4 = 16
Critical value of F at DF = 3, 16 is 2.46
Anova Summary
Source | DF | SS | MS | F |
Group | 3 | 12923.55 | 4307.85 | 4.087 |
Error | 16 | 16863.75 | 1053.984 | |
Total | 19 | 29787.3 |
Let Ti be the total energy consumed for region i, ni be number of observations of region i.
Let G be the total energy consumed of all observations and N be total number of observations.
ΣX2 is sum of squares of all observations.
T1 = 804, T2 = 440.4 , T3 = 349.3, T4 = 337.7
G = 804 + 440.4 + 349.3 + 337.7 = 1931.4
ΣX2 = 112769 + 41988.86 + 28685.83 + 32858.93 = 216302.6
SST = ΣX2 - G2/N = 216302.6 - 1931.42/20 = 29787.3
SSTR = ΣT2/n - G2/N = (8042 /6 + 440.42 /5 + 349.32 /5 + 337.72 /4) - 1931.42/20 = 12923.55
SSE = 29787.3 - 12923.55 = 16863.75
MSTR = SSTR / df for group = 12923.55 / 3 = 4307.85
MSE = SSE / df for error = 16863.75 / 16 = 1053.984
F = MSTR / MSE = 4307.85 / 1053.984 = 4.087
As, the observed value of F (4.087) is greater than the critical value (2.46), we reject the null hypothesis and conclude that the at least one average energy consumption is different among the four regions.