CHAPTER – IV

LATIN SQUARE DESIGN

4.1 INTRODUCTION

The Latin square Design is one of the maximum essential designs used in lots of
experimentation. It gives greater possibility than Complete Randomized Design and
Randomized Block Design for the Reduction of experimental mistakes through skillful planning.
In this chapter we see the Latin Square Design in the Analysis of the effectiveness of four
different traffic violations on road accident.

4.2 LATIN SQUARE DESIGN

A Latin Square design is a process of placing treatments like A, B and C, so that they show in a
balanced fashion within a square block or field.

A B C
B C A
C A B
This kind of design is used to reduce systematic error due to rows and columns.
Replicates are also included in this design.

In this design the number of treatments is equal to the number of replications. Thus in case of n
treatments, there have to be n x n = n2 (N) experimental units. The whole experimental area is
divided into n2 experimental units arranged in a square so that each row as well as each column
contains n units. The n treatments are then allocated at random of these rows and columns in
such a way that every treatment occurs once and only once in each row and in each column. This
is also known as Three Way Classification for ANOVA.

4.2.1 ADVANTAGES OF LATIN SQUARE DESIGN

 You can manipulate version in direction.

  You can boom performance compared to the randomized absolutely
block layout.
 It gives way of analyzing 3 elements on the equal time.

4.2.2 DISADVANTAGES OF LATIN SQUARE DESIGN

 The fundamental assumption that there is no interaction between the three factors
of variation may not be true in general.
 All 3-factor should be on the equal variety of levels.
4.3 ANOVA TABLE FOR LATIN SQUARE DESIGN

Source of Degree of Sum of squares Mean Sum of Variance Ratio (F-

variation freedom (SS) Squares (MS) ratio)
Between rows n-1 SS row = R1 R1 MS row
MS row = F1 =
n−1 MS error
F
Between n-1 SS column = R2 R2 MS column
MS column = F2 = F
n−1 MS error
columns
Between n-1 SS Treatment = R3 R3 MS treatment
MS treatment = F3 = F
n−1 MS error
treatments
Error (n-1) (n-2) SS error = R4 MS error =
R4
(n−1)(n−2)
Total n2 -1 SST

The various sums of squares can be obtained by using the following formulae.

1. Total sum of squares

2
SST = ∑ ∑ y ij-
2 G

i j N
Where, G = Grand total
N = Total number of experimental units

2. Between row sum of squares

1 G
2
SS row = R1 = ∑ R
2
n i i N
-

3. Between column sum of squares

1 2
∑ Cj - G
2
SS column = R2 =
n j N
 4. Between treatment sum of squares
1 2
∑ Tk - G
2
SS Treatment = R3 =
n k N

5. Error sum of squares

SS error = R4 = SST - R1 - R2 - R3

After calculation the F-ratio. Compare the calculated value of F with the table value of F for
given degree of freedom at 5% level of significance. If the calculated value of F is less than the
tabulated value we accept our null hypothesis and we may concluded that there is no significant
different between rows, columns and treatments.

4.4 ANALYSIS OF FOUR DIFFERENT TRAFFIC VIOLATION ON ROAD

ACCIDENTS USING L.S.D

The data of road Accidents with various levels from September 2021 to December 2021 in
Chennai, Madurai, Coimbatore and Thanjavur.

TABLE 1: September 2021

Drunken Over Mobile Signal

drive speed phone using jumping
Chennai 12 15 25 19
Madurai 19 9 9 13
Coimbatore 25 17 17 14
Thanjavur 12 4 4 17

TABLE 2: October 2021

Drunken Over Mobile Signal

drive speed phone using jumping
Chennai 20 23 15 12
Madurai 9 20 25 19
 Coimbatore 14 19 21 23
Thanjavur 7 13 20 17

TABLE 3: November 2021

Drunken Over Mobile Signal

drive speed phone using jumping
Chennai 20 12 19 22
Madurai 21 12 21 9
Coimbatore 7 14 21 17
Thanjavur 23 21 25 20

TABLE 4: December 2021

Drunken Over Mobile Signal

drive speed phone using jumping
Chennai 14 19 21 20
Madurai 23 18 19 25
Coimbatore 20 12 20 9
Thanjavur 15 7 4 15

TABLE 5: DATA REPRESENTATION IN LSD

Sep Oct Nov Dec

Chennai C(25) B(23) A(20) D(20)
Madurai A(19) D(19) C(21) B(18)
Coimbatore B(19) A(14) D(17) C(20)
Thanjavur D(17) C(20) B(21) A(15)
4.4.1 Calculation of an ANONA table using F-test

 Null hypothesis: There is no significant difference between rows, columns and

treatments.

Row/Column 1 2 3 4 Total
1 25 23 20 20 88
2 19 19 21 18 77
3 19 14 17 20 70
4 17 20 21 15 73
Total 80 76 79 73 308

 A- Drunken drive = 68
B- over speed = 81
C- Using Mobile while Driving = 86
D- Signal Jumping = 73
2
G
 Correction Factors (CF) =
N
2
308
=
16
94864
=
16
CF = 5929
 Total sum of squares
SST = 6042 – 5929
= 113

Where, ∑ ∑ y 2ij = 6042

i j

CF = 5929
 Between row sum of squares
1
SS row = R1 = (882 +772+702+732) - 5929
4
 = 5975.5 – 5929

= 46.5

 Between column of squares

1
SS column = R2 = (802 +762 +792+732) - 5929
4

= 5936.5 – 5929 = 7.5

 Between column of squares

1
SS Treatment = R3 = (682+812+86 2+732) - 5929
4

= 5977.5 – 5929

= 48.5

 Error sum of squares

SS error = R4 = SST - R1 - R2 - R3
= 113- 46.5-7.5-48.5
= 10.5

4.4.2 Three factor ANOVA table

Source of Degree of Sum of Mean F-Ratio F Table Value at

Variation freedom Square Square 5% level
Row 3 46.5 15.5 F 1 = 8.86 F(3,6) = 4.76
Column 3 7.5 2.5 F 2 = 1.43 F(3,6) = 4.76
Treatment 3 48.5 16.17 F 3= 9.24 F(3,6) = 4.76
Error 6 10.5 1.75
Total 15 113

Since, F1 > 4.76, there is a significant difference between rows,

F2 < 4.76, there is no significant difference between columns,

 F3 > 4.76, there is a significant difference between treatments.

4.4.2 Decision Rule:

 Standard Error (SE) =

√ EMS
r
=
√
1.75
4
= 0.66

Where, EMS = Error Mean Square

 Coefficient of Variation (CV %) =

√ EMS
y
* 100 =
√1.75
10.5
*100 = 40.825

4.4.3 TREATMENT MEAN

Treatments Drunken Over speed Mobile phone using Signal

drive jumping
Mean of 17 20.25 21.5 18.2
Treatments
S.E 0.66
C.V % 40.825
 25

20

15

10

0
Drunken drive Over Speed Mobile phone using Signal jumping

Figure 1: Treatment means comparison

Comparing the four methods with each other in the other in the bar chart mentioned above.
Comparing to these four methods to each other, the third method is more like to cause more
accidents. That is using Mobile Phone while driving is cause more accidents.

4.5 CONCLUSION

We discussed the road accidents caused by four types of factors in four different cities in four
months using Latin Square design and explore which factor caused the more accident. The use of
F-distribution test in this study was a better tool. It was used to check whether difference
between the treatments where significant. From the ANOVA table we conclude that the F-
calculated value for row, column and treatments are 8.86, 1.43 and 9.24. The section 4.4.3
represents that the treatment means. The Standard error and the coefficient of variations are 0.66
and 40.825. Therefore it is significant based on the row, column and treatment at 5% level.