Example of a Completely
Randomized Design
Refer to sections 14.2 and 14.3 of text (8th ed), 15.2 and
15.3 of 7th ed.
An experiment was conducted to compare 3 brands of gasoline. Each brand
was used with 7 different cars of the same weight and engine size, under
similar operating conditions. The data collected were the miles per gallon
obtain for each of the cars.
Brand A: 19 24 24 21 20 22 25
Brand B: 25 26 34 25 24 24 23
Brand C: 25 31 28 29 28 30 28
Do the data provide sufficient evidence to indicate a difference among
the brands of gasoline?
First, some exploratory analysis. Start with some boxplots to compare mileages
of the 3 brands.

Locations of the centre (average) of each distribution do not appear to
be the same.
Compare means, etc. among the 3 brands.
Descriptive Statistics: Mileage by Brand
# Some output deleted
Variable
Brand N Mean Median TrMean StDev
Mileage
A 7 22.143 22.000 22.143 2.268
B 7 24.286 24.000 24.286 1.113
C 7 28.429 28.000 28.429 1.902
One-way ANOVA: Mileage versus Brand
Analysis of Variance for Mileage
Source
DF SS MS F P
Brand
2 142.95 71.48
21.44 0.000
Error
18 60.00 3.33
Total
20 202.95
Individual 95% CIs For Mean
Based on Pooled StDev
Level
N Mean StDev
--------+---------+---------+--------
A
7 22.143 2.268
(-----*----)
B
7 24.286 1.113 (-----*-----)
C
7 28.429 1.902 (-----*-----)
--------+---------+---------+--------
Pooled StDev =
1.826 22.5 25.0 27.5
Test if there is a difference among the 3 brands: