STAT412

Analysis of Variance for a Randomized Block Design in Minitab

Example: An accounting firm, prior to introducing in the firm widespread training in statistical sampling for auditing, tested three
training methods: (1) study at home with programmed training materials, (2) training sessions at local offices conducted by local
staff, and (3) training session in Chicago conducted by a national staff. Thirty officers were grouped into ten blocks of three,
according to time elapsed since college graduation, and the auditors in each block were randomly assigned to the three training
methods. At the end of the training, each auditor was asked to analyze a complex case involving statistical applications; a
proficiency measure based on this analysis was obtained for each auditor. The results were (block 1 consists of auditors graduated
most recently, block 10 consists of those graduated most distantly):

Block 1 2 3
1 73 81 92
2 76 78 89
3 75 76 87
4 74 77 90
5 76 71 88
6 73 75 86
7 68 72 88
8 64 74 82
9 65 73 81
10 62 69 78

INPUTTING DATA:
TREATMENT BLOCK SCORE
1 1 73
2 1 81
3 1 92
1 2 76
2 2 78
3 2 89
. . .
. . .

COMMANDS IN MINITAB: STAT > ANOVA > General Linear Model > RESPONSE > SCORE
MODEL > TREATMENT BLOCK
COMPARISONS > Pairwise comparisons > Terms > TREATMENT > OK
GRAPHS > Histogram of residuals, normal plot of residuals,
residuals vs. fits > OK > OK
GRAPH > Scatterplot > With Groups > y-variable SCORE > X-VARIABLE
TREATMENT > Categorical Variable BLOCK > Data View > Connect Line
OK > OK > OK

ASSESSING THE REASONABLENESS OF THE NORMALITY ASSUMPTION:

Histogram of Residuals Normal Probability Plot of Residuals

12 99

95
10
90

80
8
70
Frequency

Percent

60
6 50
40
30
4 20

10
2 5

0 1
-4 -2 0 2 4 -5.0 -2.5 0.0 2.5 5.0
Residual Residual
 ASSESSING THE REASONABLENESS OF THE EQUAL VARIANCE ASSUMPTION:

Residuals Versus Fits

5.0

2.5

Residual
0.0

-2.5

-5.0
60 65 70 75 80 85 90
Fitted Value

ASSESSING THE APPROPRIATENESS OF THE NO INTERACTION ASSUMPTION:

The Scores for Each Method by Block

95 block
1
2
90
3
4
85 5
6
7
80 8
score

9
10
75

70

65

60
1 2 3
treat

Interpretation: From the appearance of the graphs, the normality and equal variance assumptions are
reasonable. There is some concern, however, about the appropriateness of the no interaction
assumption. Now we perform an analysis for a randomized block design.

Minitab Output for Performing Analysis for Randomized Block Design:

Analysis of Variance for score, using Adjusted SS for Tests

Source DF Seq SS Adj SS Adj MS F P

treat 2 1295.00 1295.00 647.50 103.75 0.000 (H0: 1.=2.=3.)
block 9 433.37 433.37 48.15 7.72 0.000
Error 18 112.33 112.33 6.24
Total 29 1840.70

S = 2.49815 R-Sq = 93.90% R-Sq(adj) = 90.17%

Unusual Observations for score

Obs score Fit SE Fit Residual St Resid

5 76.0000 71.8333 1.5800 4.1667 2.15 R
15 71.0000 75.8333 1.5800 -4.8333 -2.50 R

R denotes an observation with a large standardized residual.

H0: 1=2=3

Ha: At least two of the treatment means are different.

=.05
F=103.75
p-value=.000 < thus reject H0

Conclusion: The data provide sufficient evidence to conclude that at least two of the
treatment means are different. Now we will apply Tukeys multiple comparisons procedure to
locate which treatment means are different.

Grouping Information Using Tukey Method and 95.0% Confidence

treat N Mean Grouping

3 10 86.1 A
2 10 74.6 B
1 10 70.6 C

Means that do not share a letter are significantly different.

Tukey 95.0% Simultaneous Confidence Intervals

Response Variable score
All Pairwise Comparisons among Levels of treat
treat = 1 subtracted from:

treat Lower Center Upper --------+---------+---------+--------

2 1.148 4.000 6.852 (-----*-----)
3 12.648 15.500 18.352 (-----*-----)
--------+---------+---------+--------
5.0 10.0 15.0

treat = 2 subtracted from:

treat Lower Center Upper --------+---------+---------+--------

3 8.648 11.50 14.35 (-----*-----)
--------+---------+---------+--------
5.0 10.0 15.0