TOPIC 3: NESTED DESIGNS

3.0 Objectives
By the end of this topic, the learner should be able to:

i) Distinguish between the layout of factorial designs and nested designs

ii) Understand the necessary assumptions when analyzing of nested designs
iii) Carry out analysis of a split plot designs, split-split plot designs and two-stage
nested design

3.1 Nested Designs

Nested and Split Plot experiments are multifactor experiments that have some
important industrial applications although historically these come out of agricultural
contexts. Nested and split-plot designs frequently involve one or more random factors.
"Split plot" designs -- here we are originally talking about fields which are divided into
whole and split plots, and then individual plots get assigned different treatments. For
instance, one whole plot might have different irrigation techniques or fertilization
strategies applied, or the soil might be prepared in a different way. The whole plot
serves as the experimental unit for this particular treatment. Then we could divide each
whole plot into sub plots, and each subplot is the experimental unit for another
treatment factor.

Whenever we talk about split plot designs we focus on the experimental unit for a
particular treatment factor.
Nested designs arise when every level of one factor (B, say) is combined with (“nested
inside") only one level of another factor (A, say) in the design; this compares to the
previously discussed crossed design (or complete factorial designs) where all of the
levels of factor A are combined with all of the levels of factor B.
A1 A2 A3

B1 B2 B3 B1 B2 B3 B1 B2 B3

Page 1 of 13
Such an arrangement is called a nest or a hierarchical design with levels of factor B
nested under the levels of a factor A.
The Two-Stage Nested Design
When factor B is nested in levels of factor A, the levels of the nested factor don't have
exactly the same meaning under each level of the main factor, in this case factor A. In a
nested design, the levels of factor (B) are not identical to each other at different levels of
factor (A), although they might have the same labels. For example, if A is school and B
is teacher, teacher 1 will differ between the schools. This has to be kept in mind when
trying to determine if the design is crossed or nested. To be crossed, the same teacher
needs to teach at all the schools.

As another example, consider a company that purchases material from three suppliers
and the material comes in batches. In this case, we might have 4 batches from each
supplier, but the batches don't have the same characteristics of quality when purchased
from different suppliers. Therefore, the batches would be nested. When we have a
nested factor and you want to represent this in the model the identity of the batch
always requires an index of the factor in which it is nested.
General layout is as displayed above which can also be displayed as follows

Figure: Two Stage Nested Design

Page 2 of 13
For example:
In order to compare cleanliness in various restaurants in various cities, we take a
random sample of 3 cities in Kenya. In each city, we take a random sample of 3
restaurants and in each we take a location at random in the kitchen and take a bacteria
count.

Cities Restaurant Place

A 1 1
2
2 1
2
3 1
2
B 1 1
2
2 1
2
3 1
2
C 1 1
2
2 1
2
3 1
2
This is a two stage nested or hierarchical design with restraints nested within cities.
The model for the two stage nested design is of the form:
𝑦𝑖𝑗𝑘 = 𝜇 + 𝜏𝑖 + 𝛽𝑗 𝑖 + 𝜀 𝑖𝑗 𝑘 with 𝑖 = 1, … , 𝑎, 𝑗 = 1, … , 𝑏, 𝑘 = 1, … , 𝑟
Where
𝜇 = overall mean
𝜏𝑖 = effect of 𝑖 𝑡𝑕 factor A
𝜀 𝑖𝑗 𝑘 = random error term
Assumptions:
i) There is no interaction between A and B

Page 3 of 13
ii) The subscript 𝑗(𝑖) indicates that the 𝑗 𝑡𝑕 level of factor B is nested under the 𝑖 𝑡𝑕 each
level of factor A
iii) There are 𝑎 levels of factor 𝐴 and 𝑏 levels of factor 𝐵 nested under each level of
factor A and 𝑟 replicates.
iv) The subscript 𝑖𝑗 𝑘 is used for the error term

In the experiment of cleanliness of restraints, the model is given as follows;

𝑦𝑖𝑗𝑘 = 𝜇 + 𝜏𝑖 + 𝛽𝑗 𝑖 + 𝜀 𝑖𝑗 𝑘 with 𝑖 = 1,2,3, 𝑗 = 1,2,3, 𝑘 = 1,2
𝜀 𝑖𝑗 𝑘 ~𝑁(0, 𝜎 2 )
𝜏𝑖 ~𝑁(0, 𝜎𝐴2 )
𝛽𝑗 𝑖 ~𝑁(0, 𝜎𝐵2 )
𝑦𝑖𝑗𝑘 - Bacteria count at the 𝑘 𝑡𝑕 place in the 𝑗 𝑡𝑕 restaurant in the 𝑖 𝑡𝑕 city

𝜏𝑖 – 𝑖 𝑡𝑕 city effect
𝛽𝑗 𝑖 - effect of restaurant 𝑗 in the city
𝜀 𝑖𝑗 𝑘 - random variability effect between places in a restaurant
The sum of squares
The total sum of squares: 𝑆𝑆𝑡𝑜𝑡𝑎𝑙 = 𝑆𝑆𝐴 + 𝑆𝑆𝐵(𝐴) + 𝑆𝑆𝑒𝑟𝑟𝑜𝑟
𝑎 2
𝑖=1 𝑦… 𝑦..2
𝑆𝑆𝐴 = − , 𝑤𝑖𝑡𝑕 𝑑𝑓 (𝑎 − 1)
𝑏𝑟 𝑎𝑏𝑟

𝑎 𝑏 2 𝑎 2
𝐼=1 𝑗 =1 𝑦𝑖𝑗 . 𝑖=1 𝑦…
𝑆𝑆𝐵(𝐴) = − 𝑤𝑖𝑡𝑕 𝑑𝑓 𝑎(𝑏 − 1)
𝑟 𝑏𝑟

𝑎 𝑏 2
2 𝐼=1 𝑗 =1 𝑦𝑖𝑗 .
𝑆𝑆𝑒𝑟𝑟𝑜𝑟 = 𝑦𝑖𝑗𝑘 − 𝑤𝑖𝑡𝑕 𝑑𝑓 𝑎𝑏(𝑟 − 1)
𝑟
𝑖 𝑗 𝑘

2
𝑦..2
𝑆𝑆𝑡𝑜𝑡𝑎𝑙 = 𝑦𝑖𝑗𝑘 − 𝑤𝑖𝑡𝑕 𝑑𝑓 𝑎𝑏𝑟 − 1
𝑎𝑏𝑟
𝑖 𝑗 𝑘

ANOVA Table

Page 4 of 13
 Source of Variation Df SS MS

A 𝑎−1 𝑆𝑆𝐴 𝑀𝑆𝐴

B within A 𝑎(𝑏 − 1) 𝑆𝑆𝐵 𝑀𝑆𝐵

Error 𝑎𝑏(𝑟 − 1) 𝑆𝑆𝑒𝑟𝑟𝑜𝑟 𝑀𝑆𝑒𝑟𝑟𝑜𝑟

Total 𝑎𝑏𝑟 − 1 𝑆𝑆𝑡𝑜𝑡𝑎𝑙

𝑏𝑟 𝜏𝑖2
𝐸 𝑀𝑆𝐴 = 𝜎 2 + 𝑛𝜎𝛽2 +
(𝑎 − 1)
𝑟 𝛽𝑗2(𝑖)
𝐸 𝑀𝑆𝐵(𝐴) = 𝜎 + 2
= 𝜎 2 + 𝑛𝜎𝛽2
𝑎(𝑏 − 1)
𝐸 𝑀𝑆𝐸 = 𝜎 2
Tests of hypothesis
(i) A and B fixed 𝐻0 : 𝜏𝑖 = 0 , 𝑖 = 1, … , 𝑎 , vs 𝐻0 : 𝜏𝑖 ≠ 0 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑖
𝑴𝑺
Test statistic 𝑭 = 𝑴𝑺𝑨
𝑬

𝐻0 : 𝛽𝑗 𝑖 =0 , , vs 𝐻0 : 𝛽𝑗 𝑖 ≠0
𝑴𝑺𝑩(𝑨)
Test statistic 𝑭= 𝑴𝑺𝑬

(ii) If A is fixed and B is random then

𝐻0 : 𝜏𝑖 = 0 , 𝑖 = 1, … , 𝑎 , vs 𝐻0 : 𝜏𝑖 ≠ 0 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑖
𝑴𝑺𝑨
Test statistic 𝑭 = 𝑴𝑺
𝑩(𝑨)

(iii) To test 𝐻0 : 𝜎𝛽2 = 0 vs 𝐻0 : 𝜎𝛽2 ≠ 0

𝑴𝑺𝑨
Test statistic 𝑭 = 𝑴𝑺
𝑩(𝑨)

Example 3.1
A manufacturing engineer is studying the dimensional variability. For a particular
componenet that is produced on three machines. Each machine has two spindles and
four components are randomly selected from ach spindle. The results are given below
(both factors are fixed)

Page 5 of 13
 machine 1 2 3
spindle 1 2 1 2 1 2
12 8 14 12 14 16
9 9 15 10 10 15
11 10 13 12 12 15
12 8 14 13 11 14
Construct the ANOVA table and test the necessary hypotheses (EXERCISE)

3.2 Split plot Designs

In factorial experiments, certain factors may require larger plots than others. For
example, experiments on irrigation, tillage and others require relatively larger plots. On
the other hand plots on fertilizers, pesticides etc need smaller plots. In such situations, a
split plot can be used.

In spilt plot design the larger plots are called Main plots and the smaller plots within
the larger plots are called sub-plots. The factor levels applied to the main plot is called
main plot treatment and the factor level allocated to the sub-plots is called sub-plot
treatments. Randomization is carried out on the main plots and the sub-plots
separately.

The sub-plot factor is tested more accurately than the main plot factor (smaller error for
subplots). Therefore its best to assign the factors we are interested in to the sub-plots if
possible.

Situations in which Split –plot designs are applied:

a) If you wish to measure one factor more accurately than the other
b) The nature of the factor that necessitates that one be applied to larger plots than the
others. For example if factor A is the type of ploughing say disc ploughing, it
requires larger plots than a Factor B - type of wheat varieties (E, F, C, D).
c) Split plots designs can also be used frequently when measures are repeated several
times.

Analysis of Split-Plot Designs

The linear model for the split plot experiments in RCBD is

𝑌𝑖𝑗𝑘 = 𝜇 + 𝛼𝑖 + 𝛽𝑗 + 𝛼𝛽 𝑖𝑗 + 𝛾𝑘 + 𝛽𝛾 𝑗𝑘 + 𝜀𝑖𝑗𝑘

𝑖 = 1, … , 𝑟 𝑗 = 1, … 𝑎 𝑘 = 1, … , 𝑏

𝜇-overall mean

Page 6 of 13
𝛼𝑖 − Effect of the 𝑖 𝑡𝑕 block

𝛽𝑗 − Effect of the 𝑗 𝑡𝑕 main block factor A

𝛾𝑘 - Effect of the 𝑘 𝑡𝑕 subplot factor B

𝛼𝛽 𝑖𝑗 - main plot error or error (a)

𝛽𝛾 𝑖𝑗𝑘 – interaction effect of the 𝑗 𝑡𝑕 main plot factor A and 𝑘 𝑡𝑕 subplot factor B.

𝜀𝑖𝑗𝑘 - subplot error (error b) and 𝜀𝑖𝑗𝑘 ~𝑁(0, 𝜎 2 )

Model checking:
Since there are two error terms in the model, two kinds of residuals need to be checked,
the WP (whole plot) and SP (split plot) residuals. To calculate them, first
calculate the residuals in the usual way (response values minus fitted values). Then
• WP residuals are obtained by averaging residuals corresponding to each whole
plot.
• SP (split plot) residuals are obtained by subtracting WP residuals from residuals.
Model-checking plots:
• Normal plot of SP residuals (to check normality of split-plot errors)
• Normal plot of WP residuals (to check normality of whole-plot errors)
• Plot SP residuals against fitted values (to check constant variance for SP errors)
• Plot WP residuals against the average fitted value for the corresponding whole
plot (to check constant variance for WP errors)
• Plot WP residuals against SP residuals (to check independence of WP and SP
errors)

As suggested by the form of the model, the analysis combines two separate analyses:
the : whole plot analysis and the split-plot analysis. The idea is that the whole plots act
like blocks for the split plot analysis. The sum of squares for whole plots, is calculated in
a similar fashion to the sum of squares for blocks in a randomized complete block
design.

Page 7 of 13
The total sum of squares can be partitioned in the following components:

𝑆𝑆𝑡𝑜𝑡𝑎𝑙 = 𝑆𝑆𝐵𝑙𝑜𝑐𝑘 + 𝑆𝑆𝐴 + 𝑆𝑆𝑒𝑟𝑟𝑜𝑟 𝑎 + 𝑆𝑆𝐵 + 𝑆𝑆𝐴𝐵 + 𝑆𝑆𝑒𝑟𝑟𝑜𝑟 𝑏

Computation:
2
𝑌𝑖𝑗𝑘 𝑌…2
𝐶𝑇 = =
𝑎×𝑏×𝑟 𝑎×𝑏×𝑟

2
𝑆𝑆𝑡𝑜𝑡𝑎𝑙 = 𝑌𝑖𝑗𝑘 − 𝐶𝑇

2
𝑖 𝑌𝑖..
𝑆𝑆𝐵𝑙𝑜𝑐𝑘 = − 𝐶𝑇
𝑎𝑏
2
𝑗 𝑌.𝑗 .
𝑆𝑆𝐴 = − 𝐶𝑇
𝑏𝑟

𝑌𝑖𝑗2.
𝑆𝑆𝑚𝑎𝑖𝑛 𝑝𝑙𝑜𝑡 = − 𝐶𝑇
𝑏
𝑆𝑆𝑒𝑟𝑟𝑜𝑟 𝑎 = 𝑆𝑆𝑚𝑎𝑖𝑛 𝑝𝑙𝑜𝑡 − 𝑆𝑆𝐴 − 𝑆𝑆𝐵𝑙𝑜𝑐𝑘
2
𝑖 𝑌..𝑘
𝑆𝑆𝐵 = − 𝐶𝑇
𝑎𝑟
2
𝑌.𝑗𝑘
𝑆𝑆𝐴𝐵 = − 𝐶𝑇 − 𝑆𝑆𝐴 − 𝑆𝑆𝐵
𝑟
𝑆𝑆𝑒𝑟𝑟𝑜𝑟 𝑏 = 𝑆𝑆𝑡𝑜𝑡𝑎𝑙 − 𝑆𝑆𝑚𝑎𝑖𝑛 𝑝𝑙𝑜𝑡 − 𝑆𝑆𝐵 − 𝑆𝑆𝐴𝐵

ANOVA

Source of Df SS MS
Variation

Block 𝑟−1 𝑆𝑆𝐵𝑙𝑜𝑐𝑘

A 𝑎−1 𝑆𝑆𝐴 𝑀𝑆𝐴

Error(a) 𝑎 − 1 (𝑟 − 1) 𝑆𝑆𝑒𝑟𝑟𝑜𝑟 (𝑎) 𝑀𝑆𝑒𝑟𝑟𝑜𝑟 (𝑎)

B 𝑏−1 𝑆𝑆𝐵 𝑀𝑆𝐵

Page 8 of 13
 AB (𝑎 − 1)(𝑏 − 1) 𝑆𝑆𝐴𝐵 𝑀𝑆𝐴𝐵

Error(b) 𝑎(𝑟 − 1)(𝑏 − 1) 𝑆𝑆𝑒𝑟𝑟𝑜𝑟 (𝑏) 𝑀𝑆𝑒𝑟𝑟𝑜𝑟 (𝑏)

Total 𝑎𝑏𝑟 − 1 𝑆𝑆𝑡𝑜𝑡𝑎𝑙

To test

𝐻0 : 𝐼𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑓𝑎𝑐𝑡𝑜𝑟 𝐴 𝑎𝑛𝑑 𝑓𝑎𝑐𝑡𝑜𝑟 𝐵 𝑖𝑠 𝑛𝑜𝑡 𝑠𝑖𝑔𝑛𝑖𝑓𝑖𝑐𝑎𝑛𝑡

or 𝐻0 : 𝛽𝛾 𝑗𝑘 = 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑗 & 𝑘

vs
𝐻0 : 𝑇𝑕𝑒𝑟𝑒 𝑖𝑠 𝐼𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑓𝑎𝑐𝑡𝑜𝑟 𝐴 𝑎𝑛𝑑 𝑓𝑎𝑐𝑡𝑜𝑟 𝐵
or 𝐻0 : 𝛽𝛾 𝑗𝑘 ≠ 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑗 & 𝑘
𝑀𝑆𝐴𝐵
Test statistic is 𝐹 = 𝑀𝑆
𝑒𝑟𝑟𝑜𝑟 𝑏

Decision criteria is: we reject 𝐻0 if 𝐹𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 > 𝐹 𝑏−1 , 𝑎 𝑟−1 𝑏−1 , 𝛼

To test

𝐻0 : 𝛾𝑘 = 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑘

𝐻𝑎 : 𝛾𝑘 ≠ 0 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑘
𝑀𝑆𝐵
Test statistic is 𝐹 = 𝑀𝑆
𝑒𝑟𝑟𝑜𝑟 𝑏

Decision criteria is: we reject 𝐻0 if 𝐹𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 > 𝐹 𝑏−1 , 𝑎 𝑟−1 𝑏−1 , 𝛼

To test

𝐻0 : 𝛽𝑗 = 0 𝑓𝑜𝑟 𝑎𝑙𝑙 𝑗

𝐻𝑎 : 𝛽𝑗 ≠ 0 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑗
𝑀𝑆𝐴
Test statistic is 𝐹 = 𝑀𝑆
𝑒𝑟𝑟𝑜𝑟 𝑎

Decision criteria is: we reject 𝐻0 if 𝐹𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 > 𝐹 𝑎−1 , 𝑎−1 𝑟−1 , 𝛼

Example 3.2

Four streams of perennial rye grass were grown at each of two fertilizer levels. The four
streams were S23, N2, Kent and X. The fertilizer levels were denoted by H for heavy

Page 9 of 13
and A for average. The experiment was laid out as four blocks of four main plots for the
varieties, each split into two for the application of fertilizer. The dry matter yields were
as follows;

Blocks
Basic
Varieties Diet 1 2 3 4
S23 H 299 318 284 279
A 247 202 171 183

N2 H 315 247 289 307

A 257 175 188 174

X H 403 439 355 324

A 222 170 192 176

Kent H 382 353 383 310

A 233 216 200 143

Summary of Totals and subtotals

Blocks
Basic
Varieties Diet 1 2 3 4 Total
S23 H 299 318 284 279 1180
A 247 202 171 183 803
total 546 520 455 462
N2 H 315 247 289 307 1158
A 257 175 188 174 794
total 572 422 477 481
X H 403 439 355 324 1521
A 222 170 192 176 760
total 625 609 547 500
Kent H 382 353 383 310 1428
A 233 216 200 143 792
total 615 569 583 453

Page 10 of 13
 Total 2358 2120 2062 1896 8436

S23 N2 X KENT
H 1180 1158 1521 1428 5287
A 803 794 760 792 3149
1983 1952 2281 2220 8436

Sum of squares
84362
𝐶𝑇 = =
4×2×4
𝑆𝑆𝑡𝑜𝑡𝑎𝑙 = 2992 + 2472 + ⋯ + 1432 − 𝐶𝑇 =
23582 + 21202 + 20622 + 18962
𝑆𝑆𝐵𝑙𝑜𝑐𝑘 = − 𝐶𝑇 =
8
19832 + 19522 + 22812 + 22202
𝑆𝑆𝑠𝑡𝑟𝑎𝑖𝑛 (𝐴) = − 𝐶𝑇 =
8
𝑚𝑎𝑖𝑛 𝑝𝑙𝑜𝑡 𝑡𝑜𝑡𝑎𝑙𝑠 = 546, 572, 625, 615, … , 453
5462 + 5722 + ⋯ + 4532
𝑆𝑆𝑚𝑎𝑖𝑛 𝑝𝑙𝑜𝑡𝑠 = − 𝐶𝑇 =
2
𝑆𝑆𝑒𝑟𝑟𝑜𝑟 𝑎 = 𝑆𝑆𝑚𝑎𝑖𝑛𝑝𝑙𝑜𝑡 − 𝑆𝑆𝐵𝑙𝑜𝑐𝑘 − 𝑆𝑆𝑠𝑡𝑟𝑎𝑖𝑛 𝐴 = 7514.25
52872 + 31492
𝑆𝑆𝑚𝑎𝑛𝑢𝑟𝑒 𝐵 = − 𝐶𝑇 =
16
11802 + ⋯ + 7922
𝑆𝑆𝑆×𝑀 = 𝑆𝑆𝐴×𝐵 = − 𝐶𝑇 =
4
𝑆𝑆𝑒𝑟𝑟𝑜𝑟 𝑏 = 𝑆𝑆𝑡𝑜𝑡𝑎𝑙 − 𝑆𝑆𝑚𝑎𝑖𝑛𝑝𝑙𝑜𝑡 − 𝑆𝑆𝑚𝑎𝑛𝑢𝑟𝑒 (𝑏) − 𝑆𝑆𝑆×𝑀 = 7982.75
ANOVA
Source of df SS MS
Variation

Block 3 13718.5

Strain A 3 10303.75 3434.58

Page 11 of 13
 Error(a) 9 7514.25 𝑀𝑆𝑒𝑟𝑟𝑜𝑟 (𝑎)

Manure B 1 142845.10 𝑀𝑆𝐵

𝑆 × 𝑀(AB) 3 14435.15 𝑀𝑆𝐴𝐵

Error(b) 12 7982.75 𝑀𝑆𝑒𝑟𝑟𝑜𝑟 (𝑏)

Total 31 196793.5

To test

ℎℎ
𝐻0 : 𝑇𝑕𝑒𝑟𝑒 𝑖𝑠 𝑛𝑜 𝑖𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑠𝑡𝑟𝑎𝑖𝑛 𝑎𝑛𝑑 𝑓𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟
vs
𝐻0 : 𝑇𝑕𝑒𝑟𝑒 𝑖𝑠 𝐼𝑛𝑡𝑒𝑟𝑎𝑐𝑡𝑖𝑜𝑛 𝑏𝑒𝑡𝑤𝑒𝑒𝑛 𝑠𝑡𝑟𝑎𝑖𝑛 𝑎𝑛𝑑 𝑓𝑒𝑟𝑡𝑖𝑙𝑖𝑧𝑒𝑟
𝑀𝑆 4811 .7167
Test statistics 𝐹 = 𝑀𝑆 𝑠×𝑚 𝑏 = 665.292 = 7.23
𝑒𝑟𝑟𝑜𝑟

Decision criteria: We reject 𝐻0 if 𝐹𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 > 𝐹3,12,𝛼=0.05 .

since 𝐹𝑐𝑎𝑙𝑐𝑢𝑙𝑎𝑡𝑒𝑑 > 3.39 we reject 𝐻0 and conclude that there is interaction between the
strain and the fertilizer.
Further analysis can be done:

treatment means
XH 380.25a
kentH 357.00a
S23H 295.00b
NZH 289.5b
S23A 200.75c
NZA 198.5c
KentA 198.00c
XA 190.00c

We can also compute an LSD;

2×665.92
𝐿𝑆𝐷 = 𝑡12, 𝛼=0.05 × 𝑆𝐸𝐷 = 2.179 × = 39.74
4

Recommended treatment is 𝑠𝑡𝑟𝑎𝑖𝑛 × 𝐻 and 𝐾𝑒𝑛𝑡 × 𝐻

N/B

Page 12 of 13
Least Significant Difference
The least significant difference (LSD) test is used in the context of the analysis of
variance, when the F-ratio suggests rejection of the null hypothesis H 0, that is, when
the difference between the population means is significant. This test helps to identify
the populations whose means are statistically different.

Page 13 of 13