Lecture 10 Randomized Complete Block Design Last Lecture

BSC 311: Design and Analysis of Experiments First Semester 2019/20 Academic Year
Lecture 10: The Randomized Complete Block Design (RCBD)
Introduction
In Lecture 5, an experiment was performed in which a number of pigs were maintained on each
of the three diets.
The variable of interest was body weight gain, and the null hypothesis was that there is no
difference in mean body weight gain of pigs in the three experimental groups (i.e., on the various
diets).
In that experimental design, the Completely Randomized Design (CRD), each pig was assumed
to be independent of each other - except for the assignment of groups of pigs to the experimental
diets.
In a randomized block (RB) design, experimental units are grouped into “blocks" that are
thought to be similar.
The random assignment of units to treatments is done separately within each block.
The rationale for doing this is that, in the resulting dataset, the proportion of units receiving each
treatment is identical across blocks.
If the blocking factor is related to the outcome, then blocking can substantially increase the
precision of treatment comparisons over a completely-randomized (CR) design.
Example 1.
Consider the experiment in which each guinea pig is housed in a separate cage.
The variable in question is weight gain; the fixed effect factor to be tested is diet.
Twenty animals are to be used in this experiment, five on each of the four diets.
However, the experimenter believes there are some environmental factors that likely would
affect weight gain and does not feel that all twenty animals cages can be kept in the laboratory at
identical conditions of temperature, light, etc.
Therefore, five blocks of experimental units are established, i.e., five groups of guinea pigs. Each
block of animals consists of four guinea pigs, one on each of the experimental diets.
1
Lecture 10 Wilson Wesley Lazaro Jere
All members of a block are considered to be at identical conditions (except, of course, for diet).
The assignment of an experimental diet to each of the individuals in a block is done at random.
The table below shows the assignment of diets to animals within the blocks and the weight gain
are shown in parenthesis.
𝐻0 : The mean weight gain of guinea pigs is the same on each of four specified diets (i.e.,
𝜇1 = 𝜇2 = 𝜇3 = 𝜇4 ).
𝐻𝑎 : The mean weight gain of guinea pigs is not the same on each of four specified diets
α = 0.05
Block 1 Diet 3 (4.9) Diet 4 (8.8) Diet 1 (7.0) Diet 2 (5.3)
The preceding data (weight gains, in grams) are summarized in the table below:
Diets
Blocks 1 2 3 4 Totals: 𝐵𝑗 = ∑𝑎𝑖=1 𝑋𝑖𝑗
1 7.0 5.3 4.9 8.8 26.0
2 9.9 5.7 7.6 8.9 32.1
3 8.5 4.7 5.5 8.1 26.8
4 5.1 3.5 2.8 3.3 14.7
5 10.3 7.7 8.4 9.1 35.5
Grand total:
Totals: 𝐺𝑖 = ∑𝑏𝑗=1 𝑋𝑖𝑗 40.8 26.9 29.2 38.2 ∑𝑎𝑖=1 ∑𝑏𝑗=1 𝑋𝑖𝑗 = 135.1
Note: 𝐺𝑖 = ∑𝑏𝑗=1 𝑋𝑖𝑗 is the sum (over all groups) of all data in block j.
2
a = 4, b = 5, N = ab = 20
∑𝑎𝑖=1 ∑𝑏𝑗=1 𝑋𝑖𝑗

2
= 1011.95
1. Calculating the correction factor

2
(∑𝑎 𝑏
𝑖=1 ∑𝑗=1 𝑋𝑖𝑗 ) (135.1)2
CF = = = 912.6005
𝑁 20
2. Calculating the sums of squares and degrees of freedom
a) Total SS = ∑𝑎𝑖=1 ∑𝑏𝑗=1 𝑋𝑖𝑗

2
− 𝐶𝐹 = 1011.95 – 912.6005 = 99.3495
Total DF = 20 -1 = 19
2
∑𝑎
𝑖=1 𝐺 (40.8)2 +(26.9)2 +(29.2)2 +(38.2)2
𝑖
b) Diet SS = − 𝐶𝐹 = − 912.6005 = 27.4255
𝑏 5
Diet DF =4–1 =3
2
∑𝑏
𝑗=1 𝐵𝑗 (26.0)2 +(32.1)2 +(26.8)2 +(35.5)2
c) Block SS = − 𝐶𝐹 = − 912.6005 = 62.6470
𝑎 4
Block DF = 5 – 1 =4
d) Error SS = Total SS – Diet SS – Block SS = 99.3495 – 27.4255 – 62.6470 = 9.2770
Error DF = 19 – 3 – 4 = 12
3. Calculating the mean of squares
𝐷𝑖𝑒𝑡 𝑆𝑆 27.4255
a) Mean squares for diet = 𝐷𝑖𝑒𝑡 𝐷𝐹 = = 9.1418
3
𝐸𝑟𝑟𝑜𝑟 𝑆𝑆 9.2770
b) Mean squares for error = 𝐸𝑟𝑟𝑜𝑟 𝐷𝐹 = = 0.7731
12
3
4. ANOVA Table
Source of Sums of Degree of Mean F-value

variation Squares freedom Squares
Diets 27.4255 3 9.1418 11.8
Blocks 62.6470 4
Error 9.2770 12 0.7731
Total 99.3495 19
5. F – value and decision rule

𝑀𝑆𝐷 9.1418
F = 𝑀𝑆𝐸 = 0.7731 = 11.8
𝐹0.05(1),3,12 = 3.49
Since the calculated F is greater than the tabulated F, we reject H0. We conclude that we have
sufficient evidence that there is significant difference in mean weights of guinea pigs among the
four diets at 5 % level of significance.
Note:
 Since diet is a fixed factor and block is a random factor, there is no interaction between a
fixed and a random factor. In fact, the error in RCBD is the diet x block interaction.
 Since block is random, there is no need to test this factor. We already know that blocks are
different. If we had wanted to test the blocks, our null hypothesis would be: The variance due
to blocks is zero, i.e., 𝜎𝐵2 = 0.
4

Lecture 10 Randomized Complete Block Design Last Lecture

Uploaded by

Copyright:

Available Formats

Lecture 10 Randomized Complete Block Design Last Lecture

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture 10 Randomized Complete Block Design Last Lecture

Uploaded by

Copyright:

Available Formats

BSC 311: Design and Analysis of Experiments First Semester 2019/20 Academic Year

Lecture 10: The Randomized Complete Block Design (RCBD)

Block 1 Diet 3 (4.9) Diet 4 (8.8) Diet 1 (7.0) Diet 2 (5.3)

Block 2 Diet 1 (9.9) Diet 3 (7.6) Diet 2 (5.7) Diet 4 (8.9)

Block 3 Diet 3 (5.5) Diet 2 (4.7) Diet 4 (8.1) Diet 1 (8.5)

Block 4 Diet 4 (3.3) Diet 2 (3.5) Diet 1 (5.1) Diet 3 (2.8)

Block 5 Diet 1 (10.3) Diet 4 (9.1) Diet 3 (8.4) Diet 2 (7.7)

∑𝑎𝑖=1 ∑𝑏𝑗=1 𝑋𝑖𝑗

1. Calculating the correction factor

2. Calculating the sums of squares and degrees of freedom

a) Total SS = ∑𝑎𝑖=1 ∑𝑏𝑗=1 𝑋𝑖𝑗

d) Error SS = Total SS – Diet SS – Block SS = 99.3495 – 27.4255 – 62.6470 = 9.2770

3. Calculating the mean of squares

Source of Sums of Degree of Mean F-value

5. F – value and decision rule

You might also like