You are on page 1of 6

LATIN SQUARE DESIGN

1. Definition:
An experimental design when you have two perpendicular sources of variation. If you can block
these two sources of variation (rows x columns), you can reduce experimental error when compared
to the RCBD. Latin square is more restrictive than the RCBD. The total number of plots is the square
of the number of treatments. Each treatment appears once and only once in each row and column.

 Advantage:
- Allows the experimenter to control two sources of variation

 Disadvantages:
- The experiment becomes very large if the number of treatments is large
- The statistical analysis is complicated by missing plots and miss assigned treatments
- Error df is small if there are only a few treatments

 This limitation can be overcome by repeating a small

Latin Square and then combining the experiments.

- a 3x3 Latin Square repeated 4 times


- a 4x4 Latin Square repeated 2 times

2. Usage:
When two sources of variation must be controlled

- Slope and fertility


- Furrow irrigation and shading
- If you must plant your plots perpendicular to a linear gradient
- use only when you have more than four but fewer than ten treatments a minimum of 12 df
for error
3. Analysis data using R (or Excel) – include step by step on the analysis using the software.

Analysis data is as follow:


• Construct table of mean
• Complete an analysis of variance
• Perform significance test
• Compute mean, interaction and standard errors
• Interpret the analysis

4. Layout of experiment

5. Statistical model

Yij =  + i + j +k + ij


 = mean effect
th

β = i block effect
i
th

 =j column effect
j
th
 = k treatment effect
k

 = random error
ij

Each treatment occurs once in each block and once in each column
r=c=t
N = t2
Below is the yield (kg/ha) of four maize varieties (A, B, C, D) conducted in a Latin Square design
experiment. The row blocking was due to the gradient in water supply through irrigation system.
Also, there is a line of tree located on one side of the field that shaded the experimental area in the
morning. The line of tree is perpendicular to the gradient of water supply in this experimental plot.

Shade
Irrigation
1 2 3 4
C D B A
1
13.5 10.7 15 16.2
B A C D
2
14.1 15 13.3 10.5
D C A B
3
10.8 17.2 16.2 18.7
A B D C
4
14.6 15.3 8.9 13.2
- Design: Latin square - A-D: Maize var.
- Row: Shade
- Column: Irrigation

Ho = There is no significant difference between the effect gradient of irrigation system and shade on
the yield (kg/ha) of four maize varieties

Ha = There is significant difference between the effect gradient of irrigation system and shade on the
yield (kg/ha) of four maize varieties

6. Run appropriate statistical analysis (T-TEST, ANOVA, or regression analysis) in order to test the effect of
treatment/factors or the interaction between factors.

Table 1: The anova table of Table 3: The effect gradient of irrigation system and shade on the yield (kg/ha) of four maize varieties

Type Mean F
Source DF Pr > F
III SS Square Value

Shade 3 6.8 2.2667 5 0.0452

Irrigation 3 18.355 6.1183 13.5 0.0045


Treatment 3 78.925 26.3083 58.03 <.0001
Error 6 2.72 0.4533
Corrected 15 106.8
Total
The p-value for treatment, irrigation (column) and shade (row) are smaller than alpha, α =
0.05. Thus, there is enough evidence to reject Ho. There is significant difference between the effect gradient of irrigation
system and shade on the yield (kg/ha) of four maize varieties.
7. Present the R codes that you used for all analysis and explain the function for the codes
setwd("C:/Users/Dr. Gadget UPM/Desktop/L")
getwd()

latin <-read.csv ("latin.csv", sep = ",", header = T)


head(latin)
str(latin)
latin$Irrigation <- as.factor(latin$Irrigation)
latin$Shade <- as.factor(latin$Shade)
latin$Trt <- as.factor(latin$Trt)
str(latin)
analysis <-lm(Yield~ Irrigation+Shade+Trt, latin)
anova (analysis)

library (agricolae)
out1<- LSD.test(analysis, "Trt", p.adj = "bonferroni")
out1

par(mfrow= c(1,3))
bar.group(out1$groups, ylim=c(0,20), main = "Shade on the Yield
(kg/ha)", xlab = "Treatment" , ylab = "Mean of yield (kg/ha)",
col="light blue")

8. Show the selected output for the R code


> setwd("C:/Users/Dr. Gadget UPM/Desktop/L")
> getwd()
[1] "C:/Users/Dr. Gadget UPM/Desktop/L"
> latin <-read.csv ("latin.csv", sep = ",", header = T)
> head(latin)
Irrigation Shade Trt Yield
1 1 1 C 13.5
2 2 1 D 10.7
3 3 1 B 15.0
4 4 1 A 16.2
5 1 2 B 14.1
6 2 2 A 15.0
> str(latin)
'data.frame': 16 obs. of 4 variables:
$ Irrigation: int 1 2 3 4 1 2 3 4 1 2 ...
$ Shade : int 1 1 1 1 2 2 2 2 3 3 ...
$ Trt : Factor w/ 4 levels "A","B","C","D": 3 4 2 1 2 1 3 4 4
3 ...
$ Yield : num 13.5 10.7 15 16.2 14.1 15 13.3 10.5 10.8 17.2 ..
.
> latin$Irrigation <- as.factor(latin$Irrigation)
> latin$Shade <- as.factor(latin$Shade)
> latin$Trt <- as.factor(latin$Trt)
> str(latin)
'data.frame': 16 obs. of 4 variables:
$ Irrigation: Factor w/ 4 levels "1","2","3","4": 1 2 3 4 1 2 3 4 1
2 ...
$ Shade : Factor w/ 4 levels "1","2","3","4": 1 1 1 1 2 2 2 2 3
3 ...
$ Trt : Factor w/ 4 levels "A","B","C","D": 3 4 2 1 2 1 3 4 4
3 ...
$ Yield : num 13.5 10.7 15 16.2 14.1 15 13.3 10.5 10.8 17.2 ..
.
> analysis <-lm(Yield~ Irrigation+Shade+Trt, latin)
> anova (analysis)
Analysis of Variance Table

Response: Yield
Df Sum Sq Mean Sq F value Pr(>F)
Irrigation 3 6.800 2.2667 5.000 0.045197 *
Shade 3 18.355 6.1183 13.496 0.004469 **
Trt 3 78.925 26.3083 58.033 7.987e-05 ***
Residuals 6 2.720 0.4533
---
Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
> library (agricolae)
> out1<- LSD.test(analysis, "Trt", p.adj = "bonferroni")
> out1
$statistics
MSerror Df Mean CV t.value MSD
0.4533333 6 13.95 4.826526 3.862991 1.839151

$parameters
test p.ajusted name.t ntr alpha
Fisher-LSD bonferroni Trt 4 0.05

$means
Yield std r LCL UCL Min Max Q25 Q50 Q7
5
A 15.500 0.8246211 4 14.676247 16.32375 14.6 16.2 14.900 15.60 16.20
0
B 15.775 2.0155644 4 14.951247 16.59875 14.1 18.7 14.775 15.15 16.15
0
C 14.300 1.9373521 4 13.476247 15.12375 13.2 17.2 13.275 13.40 14.42
5
D 10.225 0.8920949 4 9.401247 11.04875 8.9 10.8 10.100 10.60 10.72
5

$comparison
NULL

$groups
Yield groups
B 15.775 a
A 15.500 a
C 14.300 a
D 10.225 b

attr(,"class")
[1] "group"
> par(mfrow= c(1,3))
> bar.group(out1$groups, ylim=c(0,20), main = "Shade on the Yield (k
g/ha)", xlab = "Treatment" , ylab = "Mean of yield (kg/ha)", col="li
ght blue")

9. Present the results in a table or figure with labels and footnotes as well as the
interpretation.
Table 2: The effect gradient of irrigation system and shade on the yield (kg/ha) of four maize varieties
Treatment Mean of yield
(kg/ha)
A 15.5 a
B 15.775 a
C 14.3 b
D 10.225 c
Means with the same letter within type of maize are not significantly different at alpha = 0.05 using LSD.

Figure 1. The effect gradient of irrigation system and shade on the yield (kg/ha) of four maize varieties. Within shade
on the yield (kh/ha), means with the same letter are significantly difference at P< 0.05 using LSD

10. Interpretation of results


 The yield (kg/ha) of four maize varieties depend on the gradient
of irrigation system and shade.
 There is significant difference between the effect gradient of
irrigation system and shade on the yield (kg/ha) of four maize
varieties
 Variety B produce the highest mean of yield than variety A, C and D.
 Variety D produce the smallest mean of yield than variety A, B, and C.
 Variety A and B does not show significant effect on yield of maize.
 Variety C and D show significant effect on yield of maize
 Mean of yield is B>A>C>D
 Blocking by irrigation effect was useful in reducing experimental error (0.453)
 Distance from shade did not appear to have a significant effect

You might also like