Sei sulla pagina 1di 5

THE GENERAL m-STAGE NESTED DESIGN

If there are m completely nested factors, then such design is referred to an m-stage nested design.
As an example, suppose a foundry wishes to investigate the hardness of two different
formulations of a metal alloy. Three heats of each alloy formulation are prepared, and two ingots
selected at random from each heat for testing with two hardness measurements made on each
ingot. The situation is illustrated in the following figure.

Alloy
1 2
Formulation

Heats 1 2 3 1 2 3

Ingots 1 2 1 2 1 2 1 2 1 2 1 2

In this experiment, heats are nested under the levels of the factor alloy formulation, and ingots
are nested under the level of the factor heats. Thus, this is a three-stage nested design with two
replicates.
The model for the general three-stage nested design is

= + + + + ; i = 1, 2, a. , j = 1, 2, b ,
k = 1, 2, c , l = 1, 2, n

DESIGNS WITH NESTED AND CROSSED FACTORS


Occasionally, in a multifactor experiment, some factors will be crossed and other factors will be
nested. Anderson and McLean (1974) and Hicks (1973) call such designs nested-factorial
designs. The statistical analysis of one such design with three factors is illustrated in the
following example.

Example
It was of interest to analyze the academic performance of students from Peshawar and DI Khan.
Three co-education schools were selected at random from Peshawar and three from DI Khan.

By Shoaib Noor
Page 1
From each school five male and five female students were selected at random. The percent marks
in their respective 5th grade exams were noted and are given the table below.

Peshawar DI Khan
Student 1 2 3 4 5 1 2 3 4 5
School 1 F 60 6363 49 70 49 55 70 43 34
55 48 54 39
M 27 30 34 54 60 23 39 40 55 54
School 2 F 95 67 84 90 76 55 48 76 48 55
M 69 47 84 56 54 45 46 54 39 50
School 3 F 65 54 87 72 59 61 57 45 28 52
M 78 45 60 56 49 35 42 53 38 44

a. Using nested design, analyze the data and state whether performance of the students from
the two cities can be considered similar.
b. Can we collapse the data and simplify the model?

We have,

Peshawar DI.Khan
Student 1 2 3 4 5 1 2 3 4 5

School 1 60 63 49 70 55 43 34 48 54 39 429
502
27 30 34 54 60 23 39 40 55 54
87 93 83 124 115 66 73 88 109 93

School 2 95 67 84 90 76 55 48 76 48 55
722 516
69 47 84 56 54 45 46 54 39 50
164 114 168 146 130 100 94 130 87 105

School 3 65 54 87 72 59 61 57 45 28 52
622 455
78 45 60 56 49 35 42 53 38 44
143 99 147 128 108 96 99 98 66 96

Total 394 306 395 398 353 262 266 316 262 294

Grand Total 1846 1400 3246

By Shoaib Noor
Page 2
Computations:
The total sum of squares is

.
= = = = =

= + + ...+
= 190851 175608.6
= 15242.4

The sum of squares of schools, and districts are computed in the usual manner that is,

.
= =

+ +
=
= 177966.7 175608.6
= 2358.1
. .. .
= =

=
= 178923.867 175608.6
= 3315.267

The Schools x Districts or SD

.. .
= = =
+ + ...+
= . .

= . .
= 181749.4 175608.6 2358.1 3315.267
= 467.45

By Shoaib Noor
Page 3
. . = . ..
= = =

=
= 180374.33 178923.86
= 1450.463

. . ..
= = = =
.
= =

= =

+
. ..
=

+ + + ++ +
= +
= 185778 180374.35 181749.4 + 178923.867
= 2578.111

Finally, the sum of squares is obtain as by subtraction as


=
= 15242.4 2358.1 3315.267 467.45 1450.463 2578.111
= 5073.7

Analysis of Variance

Source of Degrees of Sum of Mean


Variation freedom Squares Square F

Schools 2 2358.1 1179.5 6.9


District (D) 1 3315.3 3315.3 19.
S*D 2 467.45 233.72 1.382
Students * D 8 1450.5 181.30 1.07
Students * S * D 16 2578.1 161.13 0.95
Error 30 5073.0 169.10

Total 59 15242.4

(b) Significant at 5 percent.

By Shoaib Noor
Page 4
Conclusion
Since the computed value of F falls in the rejection region, therefore we may conclude that the
performance of the students from two cities is considered to be dissimilar at 5 percent of level of
significance.

By Shoaib Noor
Page 5

Potrebbero piacerti anche