BALANCED INCOMPLETE BLOCK DESIGN

STA 514 2.0: Advanced Design and Analysis of Experiments

Lecture 02

Dr. Prasansha Liyanaarachchi

Department of Statistics
University of Sri Jayewardenepura
13 January 2024
Introduction to Incomplete Block Designs
• Up to now we only considered complete block designs.

• This means we would see all treatments in each block.

• In some situations, this is not possible due to

v(physical) block size is too small
vToo expensive
vNot advisable (think of a rater having to rate 7 champagne brands)

• Example:
vAn experimenter wants to compare 2 eye-drops (“new” vs. “control”).
vEvery subject gets both treatments (one per eye; at the same time)
vAt the end, measure redness on quantitative scale in every eye.
vWhat if he wanted to test 3 different eye-drop types?

• It is still a good idea to block on subjects, but obviously it is not possible to have
complete blocks in this example.
13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 2
Introduction to Balanced Incomplete Block
Designs (BIBD)
• If all treatment comparisons are of equal importance and if we wish
to estimate each treatment effect with the same precision, then the
design needs to be balanced:

1. Each block contains the same number of treatments. (Not

including all treatments in every block makes design: Incomplete)
2. Each treatment occur the same number of times in total.
3. Each pair of treatments occur together the same number of times
in total. (Balanced)

• A design that satisfies these conditions is called Balanced

Incomplete Block Design.
13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 3
Notation
𝑎: Number of treatments.
𝑏: Number of blocks.
𝑘: Number of units in a block or Number of treatments in a block.
𝑎
ways to select treatments per each block
𝑘
𝑟: Number of times each treatment occurs in the design.
𝜆: Number of times each pair of treatments appears in the same
block.
𝑟(𝑘 − 1)
𝜆=
𝑎−1
𝜆 is an integer.
𝑁 = 𝑏𝑘 = 𝑎𝑟 ∶ Total number of observations.

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 4

Example 1: Your turn….
How many times each treatment appears in the
design?

B1 B2 B3 B4 Treatment Times appear in the design

A
B
A 1 1 0 1
C
D
B 0 1 1 1
How many times each treatment pair appears
C 1 1 1 0 together in the design?
Treatment Times appear together in the
D 1 0 1 1 Pair design
A,B
A,C
A,D
B,C
B,D
C,D
13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 5
 B1 B2 B3 B4
A 1 1 0 1
B 0 1 1 1
C 1 1 1 0
D 1 0 1 1

𝑎 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 = 4
𝑏 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑙𝑜𝑐𝑘𝑠 = 4
𝑘 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 𝑖𝑛 𝑎 𝑏𝑙𝑜𝑐𝑘 = 3
𝑟 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑖𝑚𝑒𝑠 𝑒𝑎𝑐ℎ 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 𝑜𝑐𝑐𝑢𝑟𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑑𝑒𝑠𝑖𝑔𝑛 = 3
𝜆 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑖𝑚𝑒𝑠 𝑒𝑎𝑐ℎ 𝑝𝑎𝑖𝑟 𝑜𝑓 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 𝑎𝑝𝑝𝑒𝑎𝑟𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑏𝑙𝑜𝑐𝑘 =

𝑟(𝑘 − 1) 3(3 − 1)
𝜆= = =2
𝑎−1 4−1

𝑁 = 𝑏𝑘 = 𝑎𝑟 → 𝑁 = 4×3 = 12

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 6

 Example 2
Subject (Blocks)
1 2 3 4 5 6 7 8 9 10
A B E A C D B E D C
B D A C A E C B E D
C A B D E A D C B E

𝑎 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 =
𝑏 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑏𝑙𝑜𝑐𝑘𝑠 =
𝑘 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 𝑖𝑛 𝑎 𝑏𝑙𝑜𝑐𝑘 =
𝑟 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑖𝑚𝑒𝑠 𝑒𝑎𝑐ℎ 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡 𝑜𝑐𝑐𝑢𝑟𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑑𝑒𝑠𝑖𝑔𝑛 =
𝜆 = 𝑁𝑢𝑚𝑏𝑒𝑟 𝑜𝑓 𝑡𝑖𝑚𝑒𝑠 𝑒𝑎𝑐ℎ 𝑝𝑎𝑖𝑟 𝑜𝑓 𝑡𝑟𝑒𝑎𝑡𝑚𝑒𝑛𝑡𝑠 𝑎𝑝𝑝𝑒𝑎𝑟𝑠 𝑖𝑛 𝑡ℎ𝑒 𝑠𝑎𝑚𝑒 𝑏𝑙𝑜𝑐𝑘 =

𝑟(𝑘 − 1) 6(3 − 1)
𝜆= = =3
𝑎−1 5−1

𝑁 = 𝑏𝑘 = 𝑎𝑟 → 𝑁 = 6 = 30
13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 7
Statistical Model for a BIBD
𝑖 = 1,2, … , 𝑎
𝑦!" = 𝜇 + 𝜏! + 𝛽" + 𝜖!" J
𝑗 = 1,2, … , 𝑏

𝑦!" : observation for the i#$ treatment and j#$ block 𝜏! : i#$ treatment effect

𝜇: Overall mean 𝛽" : j#$ block effect

𝜖!" : Random error

Assumptions and Constraints:

𝜖!"#$ ~𝑁𝐼𝐷(0, 𝜎 % )
$ &
/ 𝜏! = / 𝛽% = 0
!"# %"#

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 8

ANOVA for BIBD
All the treatments are not presented in all the blocks. Therefore, treatment sum of
squares should be adjusted accordingly.

The total variability may be partitioned into:

𝑺𝑺𝑻 = 𝑺𝑺𝑻𝒓𝒆𝒂𝒕𝒎𝒆𝒕𝒔(𝑨𝒅𝒋𝒖𝒔𝒕𝒆𝒅) + 𝑺𝑺𝑩𝒍𝒐𝒄𝒌𝒔 + 𝑺𝑺𝑬

with respective degrees of freedom

𝑵−𝟏= 𝒂−𝟏 + 𝒃−𝟏 + 𝑵−𝒂−𝒃+𝟏 𝒘𝒉𝒆𝒓𝒆 𝑵 = 𝒃𝒌 = 𝒂𝒓

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 9

Hypothesis for treatment effect

𝐻B: 𝜏# = 𝜏C = ⋯ = 𝜏$ = 0 𝑣𝑠 𝐻#: 𝜏! ≠ 0 𝑎𝑡 𝑙𝑒𝑎𝑠𝑡 𝑜𝑛𝑒 𝑖

The appropriate statistic for testing for no differences in treatment

means is

𝑀𝑆DEF$GHFIGJ(LM%NJGFM)
𝐹B =
𝑀𝑆P

under the null hypothesis 𝐻B ⟹ 𝐹B~𝐹$Q#,SQ$Q&T#

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 10

 ANOVA Table
Source of Sum of DF Mean F p-value
Variation Squares (SS) Square (MS)
Treatments 𝑆𝑆%&'((*+") 𝑎−1 𝑀𝑆%&'( (*+") 𝐹. 𝑃(𝐹 ≥ 𝐹. )
(Adjusted) 𝑀𝑆%&'( (*+")
=
𝑀𝑆/
Blocks 𝑆𝑆01234( 𝑏−1 𝑀𝑆01234(

Error 𝑆𝑆/ 𝑁−𝑎−𝑏+1 𝑀𝑆/

Total 𝑆𝑆% 𝑁−1

Here 𝑁 = 𝑏𝑘 = 𝑎𝑟

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 11

 ANOVA - Computing formulas
The adjusted Treatment Sum of Squares

0
𝑘
𝑆𝑆&'()(+,") = 1 𝑄!%
𝜆𝑎
!./
where 𝑄! be the adjusted total for the 𝑖 (1 treatment,

3
1
𝑄! = 𝑦!. − 1 𝑛!" 𝑦." ; 𝑖 = 1,2, … , 𝑎
𝑘
"./

1 if treatment 𝑖 appears in the block

with 𝑛!" = ;
0 Otherwise

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 12

The Block Sum of Squares

3 %
1 𝑦..
𝑆𝑆4$56#) = 1 𝑦."% −
𝑘 𝑁
"./

The Total Sum of Squares

0 3 %
% 𝑦..
𝑆𝑆& = 1 1 𝑦!" −
𝑁
!./ "./

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 13

Example 03- Catalyst Experiment
Suppose that a chemical engineer thinks that the time of reaction
for a chemical process is a function of the type of catalyst
employed. Four catalysts are currently being investigated. The
experimental procedure consists of selecting a batch of raw
material, loading the pilot plant, applying each catalyst in a
separate run of the pilot plant, and observing the reaction time.
Because variations in the batches of raw material may affect the
performance of the catalysts, the engineer decides to use batches
of raw material as blocks.
Batch of Raw Material
Catalyst 1 2 3 4
1 73 74 -- 71
2 -- 75 67 72
3 73 75 68 --
4 75 -- 72 75
13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 14
• Objectives: To check whether the time of reaction for a chemical
process is a function of the type of catalyst employed.

• Response variable: Time of reaction

• Treatments: Type of catalyst: 1, 2, 3, 4

• Blocking variables: Batches of raw material: 1, 2, 3, 4

• Design: BIBD

𝑎 =4, 𝑏 =4, 𝑘 =3, 𝑟=3

𝑟(𝑘 − 1) 3(3 − 1)
𝜆= = =2
𝑎−1 4−1

𝑁 = 𝑏𝑘 = 𝑎𝑟 → 𝑁 = 4×3 = 12
13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 15
Result (Minitab Version 16)

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 16

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 17
In our example…
Ø Residuals are normally distributed (points are around a straight

line in the normal probability plot).

Ø Mean of residuals= 0 (points are scattered around 𝑒!% = 0).

Ø Variances of residuals are almost constant. (no pattern in the

points of the residual plot)

Ø If all the assumptions are satisfied, we can interpret the analysis.

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 18

STEP 1: Write down the model, define notations and the assumptions of the
model.

𝑖 = 1,2,3,4
𝑦!" = 𝜇 + 𝜏! + 𝛽" + 𝜖!" J
𝑗 = 1,2,3,4

𝑦!" : 𝑙The reaction time of the 𝑖 '5 catalyst and 𝑗'5 batch of raw material
for substrate material 𝑗 and day 𝑘
𝜇: Overall mean 𝛽" : j#$ batch of raw material effect k #$ day effect

𝜏! : i#$ catalyst effect 𝜖!" : Random error

Assumptions and Constraints:

𝜖!" ~𝑁𝐼𝐷(0, 𝜎 6 )
9 9
\ 𝜏! = \ 𝛽" = 0
!78 "78
13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 19
STEP 2: Write down the hypothesis for treatment effect, construct the ANOVA
and make conclusion.

𝐻: : 𝜏8 = 𝜏6 = 𝜏; = 𝜏9 = 0 𝑣𝑠 𝐻8 : 𝜏! ≠ 0 𝑎𝑡 𝑙𝑒𝑎𝑠𝑡 𝑜𝑛𝑒 𝑖; 𝑖 = 1,2,3,4

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 20

The 𝑝 − 𝑣𝑎𝑙𝑢𝑒 = 0.011 < 𝛼 = 0.05 (𝑜𝑟 𝐹: = 11.67 > 𝐹<=!>!<?@ = 5.41),
Therefore 𝐻: is rejected at 5% significance level. We have enough evidence
to conclude that the effect of different type of catalyst employed is
significant on the time of reaction for a chemical process.

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 21

Since, 𝐻: is rejected, multiple comparisons are needed to perform.

The output mentioned that “Means that do not share a letter are
significantly different.

According to that, Catalyst 4 is different from Catalyst 1, Catalyst 2, and

Catalyst 3.

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 22

Example 04- Gasoline Additives
An engineer is studying the mileage performance characteristics of
five types of gasoline additives. In the road test he wishes to use
cars as blocks; however, because of a time constraint, he must use
an incomplete block design. He runs the balanced design with the
five blocks that follow. Analyze the data from this experiment (use
𝛼 = 0.05) and draw conclusions.

13/01/2024 BALANCED INCOMPLETE BLOCK DESIGN 23