9 II February 2021

[Link]
 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

Measure of Modified Rotatability for Second Order Response

Surface Designs using Pairwise Balanced Designs
Jyostna P1, Victorbabu B Re2
1, 2
Department of Statistics, Acharya Nagarjuna University, Guntur-522510, India

Abstract: In this paper, measure of modified rotatability for second order response surface designs using pairwise balanced
design is suggested which enables us to assess the degree of modified rotatability for a given response surface design. It is
observed that this method sometimes leads to designs with lesser number of design points.
Keywords and phrases: Response surface designs, modified rotatable designs, pairwise balanced designs, measure of rotatability
for second order response surface designs.

I. INTRODUCTION
Response surface methodology is a statistical technique that is very useful in design and analysis of scientific experiments. In many
experimental situations, the experimenter is concerned with explaining certain aspects of a functional relationship
Y =f(x1 ,x 2 ,..., x v )+ε, where Y is the response; x1 ,x 2 ,..., x v are the levels of v-quantitative variables or factors; and ε is
the random error. Response surface methods are useful where several independent variables influence a dependent variable. The
independent variables are assumed to be continuous and controlled by the experimenter. The response is assumed to be as random
variable. For example, if a chemical engineer wishes to find the temperature (x1 ) and pressure (x 2 ) that maximizes the yield
(response) of his process, the observed response Y may be written as a function of the levels of the temperature (x1 ) and pressure
(x 2 ) as Y=f(x1 ,x 2 )+ε.
The concept of rotatability, which is very important in response surface designs, was proposed by Box and Hunter (1957). Das and
Narasimham (1962) constructed rotatable designs through balanced incomplete block designs (BIBD). Tyagi (1964) constructed
second order rotatable designs (SORD) using pairwise balanced designs (PBD). If the circumstances are such that exact rotatability
is unattainable, it is still a good idea to make the design nearly rotatable. Thus, it is important of know if a particular design is
rotatable or, if is not, to know how rotatable it is. Park et al. (1993) introduced measure of rotatability for second order response
surface designs. Victorbabu and Surekha (2012, 2013, 2015) studied measure of rotatability for second order response surface
designs using CCD, incomplete block designs and BIBD respectively. Das et al. (1999) studied response surface designs,
symmetrical and asymmetrical, rotatable and modified. Victorbabu and Vasundharadevi (2005) studied modified second order
response surface designs, rotatable designs using BIBD. Victorbabu et al. (2006) suggested modified second order response surface
designs, rotatable designs using pairwise balanced designs. Victorbabu et al. (2008) suggested modified rotatable central composite
designs. Victorbabu and Vasundharadevi (2008) studied modified second order response surface designs, rotatable designs using
symmetrical unequal block arrangements (SUBA) with two unequal block sizes. Jyostna et al. (2020) constructed measure of
modified rotatability for second order response surface designs. Jyostna and Victorbabu (2020) studied measure of modified
rotatability for second order response surface designs using BIBD.

II. CONDITIONS FOR SECOND ORDER ROTATABLE DESIGNS

Suppose we want to use the second order response surface design D= (( x iu )) to fit the surface,
v v
Yu  b   b x   b x 2    b x x  eu (1)
0 i1 i iu i1 ii iu i  j ij iu ju

where x iu denotes the level of the ith factor (i =1,2,…,v) in the uth run (u=1,2,…,N) of the experiment, e u ' s are uncorrelated
2
random errors with mean zero and variance σ . D is said to be second order rotatable design (SORD), if the variance of the
estimate of first order partial derivative of Yu (x1 ,x 2 ,...,x v ) with respect to each of independent variables (x i ) is only a function

©IJRASET: All Rights are Reserved 506

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

v
2 2
of the distance ( d = x
i=1
i ) of the point (x1,x2, …,xv) from the origin (centre) of the design. Such a spherical variance function for

estimation of second order response surface is achieved if the design points satisfy the following conditions [cf. Box and Hunter
(1957), Das and Narasimham (1962)].

1) x iu =0 , x iu x ju =0 , x iu x 2j u =0 , x iu x ju x ku =0 , x 3
iu =0 , x iu x 3ju =0 , x iu
2
x ju x ku =0 ,

x iu x ju x ku x lu =0 ; for i  j  k  l ; (2)
2
2) (i) x iu = constant = Nλ 2 ;
4
3) x iu = constant = cNλ 4 ; for all i (3)
2
4) x iu x 2ju = constant = Nλ 4 ; for i  j (4)
4
5) x iu =c x iu2 x 2ju (5)

λ4 v
6) 2
 (6)
λ 2 (c+v-1)
where c, λ 2 and λ 4 are constants and the summation is over the design points.
If the above mentioned conditions are satisfied, the variances and covariances of the estimated parameters become,
λ 4 (c+v-1)σ 2
V(bˆ 0 ) = ,
N λ 4 (c+v-1)-vλ 22 
σ2
V(bˆ i )= ,
Nλ 2
σ2
V(bˆ ij )= ,
Nλ 4
σ2  λ 4 (c+v-2)-(v-1)λ 22 
V(bˆ ii )= ,
(c-1)Nλ 4  λ 4 (c+v-1)-vλ 22 
-λ 2 σ 2
C o v ( bˆ 0 ,bˆ ii )= ,
N [λ 4 (c + v -1 )-v λ 22 ]

( λ 22 - λ 4 ) σ 2 (7)
C o v ( bˆ i i , bˆ j j ) =
( c - 1 ) N λ 4 [ λ 4 ( c + v - 1 ) - v λ 22 ]
and other covariances are zero.

III. CONDITIONS FOR MODIFIED SECOND ORDER ROTATABLE DESIGNS

Let (v, b, r, k1 , k 2 ,...k p , λ ), be an equi-replicated PBD and k= sup(k1,k2,…, kp). Then 2t(k) denotes a resolution V fractional

factorial of 2k in 1 levels, such that no interaction with less than five factors is confounded. n 0 denotes the number of central
points in the design. Let [1  (v, b, r, k1 , k 2 ,...k p , λ )] denotes the design points generated from the transpose of incidence matrix

of PBD, [1  (v, b, r, k1 , k 2 ,...k p , λ )]2

t(k)
are the b2t(k) design points generated from PBD by multiplication (cf. Raghavarao,

©IJRASET: All Rights are Reserved 507

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

1
1971). (β, 0, 0,..., 0)2 denotes the design points generated from (β, 0, 0,..., 0) point set, and U denotes combination of the
design points generated from different sets of points. The usual method of construction of SORD is to take combinations with
v
unknown constants, associate a 2 factorial combinations or a suitable fraction of it with factors each at 1 levels to make the level
codes equidistant.
All such combinations form a design. Generally, SORD need at least five levels (suitably coded) at 0, ±1, ±β for all factors
( (0, 0,...0) )- chosen centre of the design, unknown level ‘β’ are to be chosen suitably to satisfy the conditions of the rotatability)
generation of design points this way ensures satisfaction of all the conditions even though the design points contain unknown levels.
2 4 2
Alternatively, by putting some restrictions indicating some relation among  x , x
iu iu and x iu x 2ju some equations
involving the unknowns are obtained and their solution gives the unknown levels. In SORD the restriction used is
4
x iu =3 x iu2 x 2ju , i.e., c=3 . Other restrictions are also possible through, it seems, not exploited well. Das et al (1999)
2 2
proposed the restriction ( x iu) =N  x iu2 x 2ju i.e., λ 22 =λ 4 to get another series of symmetrical second order response surface
designs, which provide more precise estimates of response at specific points of interest than what is available from the
corresponding existing designs. Further, the variances and covariances of the estimated parameters are,

 (c+v-1)σ 2
V(b 0 ) 
N(c-1)

 σ2
V(bi ) 
N λ4

 σ2
V( bij ) 
Nλ 4

 2
V( bii )=
(c-1)Nλ 4

   2
Cov( b0 , bii ) = (8)
N λ 4 (c-1)

and other covariances are zero. These modifications of the variances and covariances affect the variance of the estimated response at
specific points considerably.

Using these variances and covariances, variance of estimated response at any point can be obtained. Let y u denote the estimated
response at the point (x1u , x 2u ,...x vu ) . Then,
     
V( Y u ) = V( b0 )+d 2 [V(bi )  2cov( b0 , bii )]  d 4 V( b ii )  ( x iu2 x 2ju )[ (c-3)σ 2 (c-1)Nλ 4 ]

Construction of modified response surface designs is the same as for SORD except that instead of taking c=3 the restriction
( x iu2 )2 =N  x iu2 x 2ju is to be used and this condition will provide different values of the unknowns involved. (cf. Das et al.
1999).

©IJRASET: All Rights are Reserved 508

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

IV. CONDITIONS FOR MEASURE OF ROTATABILITY FOR SECOND ORDER RESPONSE SURFACE DESIGNS
Following Box and Hunter (1957), Das and Narasimham (1962), Park et al (1993), conditions (2) to (6) and (7) give the necessary
and sufficient conditions for a measure of rotatability for any general second order response surface designs. Further we have,

V(bi ) are equal for i ,

V(bii ) are equal for i ,

V(bij ) are equal for i, j , where i  j ,

Cov(bi ,bii )=Cov(b i ,b ij )=Cov(b ii ,b ij )=Cov(b ij ,bil )=0 for all i  j, j  l, l  i . (9)

Park et al. (1993) suggested that if the conditions in (2) to (6) together with (7) and (9) are met, then the following measure
(Pv (D)) given below can be used to assess the degree of rotatability for any general second order response surface design (cf. Park
et al., 1993).

1
Pv (D) = , (10)
1+R v (D)

where
2
    

2 6v V(b ij )  2cov ( b ii , b jj )  2V(b ii )
N   (v-1)
R v (D) =  2  (11)
  (v+2)2 (v+4)(v+6)(v+8)g8

and g is the scaling factor.

   
2
On simplification, numerator of (11), [V(bij )  2 cov ( bii , b jj )  2V(b ii )] using (7) becomes (c-3)σ (c-1)Nλ 4 . Thus
R v (D) becomes
2
N  6v[(c-3)σ 2 ]2 (v-1) 
R v (D) =  2   2 2 8  (12)
 σ   [(c-1)Nλ 4 ] (v+2) (v+4)(v+6)(v+8)g 

Note: For SORD, we have c = 3 . Substituting the value of 'c' in (12) and on simplification we get R v (D) is zero. Hence from
(10), we get Pv (D) is one if and only if a design is rotatable and less than one then it is nearly rotatable design.

V. MODIFIED ROTATABILITY FOR SECOND ORDER RESPONSE SURFACE DESIGNS USING PBD (CF.
VICTORBABU ET AL (2006))
The arrangement of v-treatments in b blocks will be called a PBD of index λ and type (v, k1, k2, …, km) if each block contains
k1,k2,…km that are all distinct treatments occurs in exactly λ blocks of the design. If bi is the number of blocks of size ki = (i = 1,
m m
2,…,m), then b =  bi and λv(v-1)   bi ki (k i -1).
i=1 i=1

Let (v, b, r, k1 , k 2 ,...k p , λ ), be a PBD. Then 2t(k) denotes a resolution V fractional factorial of 2k in 1 levels, such that no

interaction with less than five factors is confounded. Let [1  (v, b, r, k1 , k 2 ,...k p , λ )] denotes the design points generated from

©IJRASET: All Rights are Reserved 509

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

the transpose of incidence matrix of PBD, [1  (v, b, r, k1 , k 2 ,...k p , λ )]2

t(k)
are the b2t(k) design points generated from PBD by

multiplication (cf. Raghavarao, 1971). Repeat these design points y1 times. (β, 0, 0,..., 0)21 denotes the design points generated
from (β, 0, 0,..., 0) point set, and repeat this set of additional design points say y 2 times and n 0 denotes the number of central
points.

The design points, y1[1-(v,b,r,k,λ)]2t(k) U y 2 (β,0,0,....0)21 U (n 0 ) will give a v dimensional modified SORD in
(y1r2t(k)  2y 2β 2 ) 2
N design points if,
y1λ2t(k)
(3λ-r)y1 2t(k)-1
β4 = ,
y2

(y1r2 t(k) +2y 2β 2 ) 2

n0 = t(k)
-[y1b2 t(k) +y 2 2v] and n 0 turns out to be an integer.
y1λ2

VI. MEASURE OF ROTATABILITY FOR SECOND ORDER RESPONSE SURFACE DESIGNS USING PBD (CF.
VICTORBABU AND SUREKHA (2013))
Let (v, b, r, k1 , k 2 ,...k p , λ ) denote a PBD. For the design points, y1[1-(v,b,r,k,λ)]2t(k) U y 2 (β,0,0,....0)21 U (n 0 ) will
give a measure of rotatability for second order response surface designs using PBD in N= y1b2t(k) +2vy 2 +n 0 design points with
r2 t(k) y1  2y 2β 4
level ‘β’ prefixed and c = .
λ2t(k) y1
We can obtain the measure of rotatability values for second order response surface designs using PBD. We have

2
 (c-3)  6v(v-1)
R v (D) =   2 2 8
 (c-1)  λ 4 (v+2) (v+4)(v+6)(v+8)g
where

1 2 t(k)-1 (b-r)y1
 , if β< +v
β y2

g= 1 2 t(k)-1 (b-r)y1
 t(k)-1 , if β > +v
 2 (b-r)y1
y 2
 +v
 y 2

1
Pv (D) =
1+R v (D)

If Pv (D) is 1 if and only if the design is rotatable, and it is smaller than one for a non-rotatable designs.

©IJRASET: All Rights are Reserved 510

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

VII. MEASURE OF MODIFIED ROTATABILITY FOR SECOND ORDER RESPONSE SURFACE DESIGNS USING
PAIRWISE BALANCED DESIGN
In this section the proposed new method of measure of modified rotatability for second order response surface designs is suggested
below:

Let (v, b, r, k1 , k 2 ,...k p , λ ), be a PBD. 2t(k) denotes a resolution V fractional factorial of 2k in 1 levels, such that no

interaction with less than five factors is confounded. [1  (v, b, r, k1 , k 2 ,...k p , λ )] denotes the design points generated from the

transpose of incidence matrix of PBD, [1  (v, b, r, k1 , k 2 ,...k p , λ )]2

t(k)
are the b2t(k) design points generated from PBD by

multiplication. Repeat these design points y1 times. Let (β, 0, 0,..., 0)21 denote the design points generated from (β, 0, 0,..., 0)
point set. Repeat this set of additional design points say y 2 times and n 0 is the number of central points.
Consider the design points, y1[1-(v,b,r,k,λ)]2t(k) U y 2 (β,0,0,....0)21 U (n 0 ) generated from PBD, we have,
2
x iu = y1r2 t(k) +y 2 2β 2  Nλ 2 (13)

4
x iu = y1r2t(k) +y 2 2β 4  cNλ 4 (14)

2
x iu x 2ju = y1λ2 t(k)  Nλ4 (15) To make

the design rotatable, we take c = 3 . From equations (14) and (15), we have

y1 (3λ-r)2 t(k)-1
β4 = ,
y2
2 2
The modified condition ( x iu ) =N  x iu2 x 2ju leads to N which is given by

(y1r2t(k)  y 2 2β 2 ) 2 t(k)
N t(k)
alternatively N may be obtained directly as y1b2 +y 2 2v+n 0 , where n 0 is given by
y1λ2
(y1r2 t(k) +2β 2 y 2 ) 2
n0 = t(k)
-[y1b2t(k) +2y 2 v] and n 0 turns out to be an integer. From equations (13) and (15) and on
y1λ2
t(k)
y r2t(k) +2y 2β 2 y λ2
simplification we get λ 2 = 1 and λ 4 = 1 .
N N
To obtain measure of modified rotatability for second order response surface designs using PBD, we have

1
Pv (D) =
1+R v (D)

2
 (c-3)  6v(v-1)
R v (D) =   2 2 8 ,
 (c-1)  λ 4 (v+2) (v+4)(v+6)(v+8)g

©IJRASET: All Rights are Reserved 511

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

where g is a scaling factor

1 y 2 t(k)-1 (b-r)
 , if β< 1 +v
 β y 2

g= 1
 otherwise
t(k)-1
 y1 2 (b-r) + v
 y2
The following table gives the values of a measure of modified rotatability for second order response surface designs using PBD. It
can be verified that Pv (D) is 1 if and only if the design is modified rotatable, and it is smaller than one for nearly modified
rotatable designs.

Example: We illustrate the measure of modified rotatability for second order response surface designs for v=10 factors with the
help of a PBD (v=10,b=11,r=5,k1 =5,k 2 =4,λ=2) The design points,

y1[1-(v=10,b=11,r=5,k1 =5,k 2 =4,λ=2)]24 U y 2 (  β,0,0,....0)21 U (n 0 ) will give a measure of modified rotatability for
second order response surface designs in N=242 design points. From (13), (14) and (15), we have

2
x iu = y1 80+y 2 2β2  Nλ 2 (16)

4
x iu = y180+y 2 2β 4  cNλ 4 (17)

2
x iu x 2ju = y1 32  Nλ 4 (18)

4 2
From equations (17) and (18) with rotatability value c=3 , y1 =1 and y 2 =2 , we get a = 4  a = 2  a=1.414214 . From
2 2
equations (16) and (18) using the modified condition with (λ 2 =λ 4 ) along with a = 2 , y1 =1 and y 2 =2 , we get N=242 ,
n 0 = 26 . For modified SORD we get Pv (D)=1 by taking a=1.414214 and scaling factor g=0.7071 . Then the design is
modified SORD using PBD.
Instead of taking a=1.414214 if we take a= 2.5 for the above PBD (v=10,b=11,r=5,k1 =5,k 2 =4,λ=2) from equations (17)
and (18), we get c=7.3828 . The scaling factor g=0.4 , R v (D)= 38.2687 and Pv (D)=0.0255 . Here Pv (D) becomes
smaller it deviates from modified rotatability.
Here we may point out that this measure of modified rotatability for second order response surface designs using PBD for v=9 and
v=10 factors has only 242 design points, whereas the corresponding measure of modified rotatability for second order response
surface design using BIBD obtained by Jyostna and Victorbabu (2020) need 726 and 441 design points respectively. Thus the new
method leads to a 9-factor and 10-factor measure of modified second order response surface designs using PBD in less number of
design points than the corresponding measure of modified rotatability for second order response surface designs using BIBD.
Table 1 gives the values of measure of modified rotatability Pv (D) for second order response surface designs using PBD, at
different values of ‘β’ for 9  v  12 . It can be verified that Pv (D) is 1, if and only if a design ‘D’ is modified rotatable. Pv (D)
becomes smaller as ‘D’ deviates from a modified rotatable design.

©IJRASET: All Rights are Reserved 512

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

Table1. Measure of modified rotatability for second order response surface designs using PBD
(9,11,5,5, 4,3, 2), N  242,β  1.414214, y1  1, y 2  2, n 0  30
β c g R v (D) Pv (D)
1.0 2.625 1 3.2802 0.9967
1.3 2.85701 0.76923 0.003 0.997
*1.414214 3 0.70711 0 1
1.6 3.3192 0.625 0.0316 0.9693
1.9 4.129 0.5263 1.3619 0.4234
2.2 5.4282 0.4545 10.1634 0.0896
2.5 7.3828 0.4 44.3147 0.0221
2.8 10.1832 0.3571 142.3824 0.007
3.1 14.044 0.3226 376.5976 0.0026

(10,11,5,5, 4, 2), N  242,β  1.414214, y1  1, y 2  2, n 0  26

β c g R v (D) Pv (D)
1.0 2.625 1 0.0028 0.9972
1.3 2.85701 0.76923 0.0026 0.9974
*1.414214 3 0.70711 0 1
1.6 3.3192 0.625 0.0433 0.9585
1.9 4.129 0.5263 1.1761 0.4595
2.2 5.4282 0.4545 105.3227 0.0094
2.5 7.3828 0.4 38.2687 0.0255
2.8 10.1832 0.3571 122.9573 0.0081
3.1 14.044 0.3226 325.2123 0.0031

(13,15,7,7,6,5,3), N  3364,β  2.828427, y1  3, y 2  3, n 0  406

β c g R v (D) Pv (D)
1.0 2.3438 1 0.005 0.995

1.3 2.3631 0.7692 0.0372 0.9641

1.6 2.4016 0.625 0.1638 0.8593

1.9 2.4691 0.5263 0.464 0.6831

2.2 2.5774 0.4545 0.8242 0.5482

2.5 2.7402 0.4 0.7112 0.5844

2.8 2.9736 0.3571 0.2121 0.825

*2.828427 3 0.3536 0 1

3.1 3.2953 0.3226 2.9537 0.2529

©IJRASET: All Rights are Reserved 513

 International Journal for Research in Applied Science & Engineering Technology (IJRASET)
ISSN: 2321-9653; IC Value: 45.98; SJ Impact Factor: 7.429
Volume 9 Issue II Feb 2021- Available at [Link]

(14,15,7,7,6,3), N  3364, β  2.828427, y1  3, y 2  3, n 0  400

β c g R v (D) Pv (D)
1.0 2.3438 1 0.0044 0.9956
1.3 2.3631 0.7692 0.0327 0.9683
1.6 2.4016 0.625 0.1438 0.8743
1.9 2.4691 0.5263 0.4075 0.7105
2.2 2.5774 0.4545 0.7238 0.5801
2.5 2.7402 0.4 0.6245 0.6156
2.8 2.9736 0.3571 0.0164 0.9838
*2.828427 3 0.3536 0 1
3.1 3.2953 0.3226 2.5939 0.2783
*indicates modified rotatability value using BIBD. (cf. Victorbabu et al. (2006))

REFERENCES
[1] Box, G.E.P. and Hunter, J.S. (1957), Multifactor experimental designs for exploring response surfaces, Annals of Mathematical Statistics, 28, 195-241.
[2] Das, M.N. and Narasimham, V.L. (1962), Construction of rotatable designs through balanced incomplete block designs, Annals of Mathematical Statistics, 33,
1421-1439.
[3] Das, M.N., Rajendra P. and Manocha, VP. (1999), Response surface designs, symmetrical and asymmetrical, rotatable and modified, Statistics and
Applications, 1, 17-34.
[4] Park, S.H, J. H. Lim and Y. Baba (1993), A measure of rotatability for second order response surface designs, Annals of Institute of Statistical Mathematics, 45,
655-664.
[5] Raghavarao, D. Constructions and Combinatorial Problems in Design of Experiments. New York: Wiley; 1971.
[6] Tyagi, B. N., (1964). On the construction of second order and third order rotatable designs through pairwise balanced designs and doubly balanced designs, Cal.
Stat. Assn Bull. 13, 150-162.
[7] Jyostna, P. Sulochana, B and Victorbabu, B. Re. (2021). Measure of modified rotatability for second order response surface designs, Journal of Mathematical
and Computational Sciences, 11 (1), 494-519.
[8] Jyostna, P. and Victorbabu, B. Re. (2020). Measure of modified rotatability for second order response surface designs using balanced incomplete block designs,
Paper communicated in Asian Journal of Probability and Statistics.
[9] Victorbabu, B. Re. and Surekha, Ch.V.V.S. (2012). Construction of measure of second order rotatable designs using central composite designs, International
Journal of Agricultural and Statistical Sciences, 8(1), 1-6.
[10] Victorbabu, B. Re., and Surekha, Ch. V. V. S. (2015). A note on measure of rotatability for second order response surface designs using balanced incomplete
block designs, Thailand Statistician, 13, 97-110.
[11] Victorbabu, B. Re. and Surekha, Ch. V. V. S. (2013). A note on measure of rotatability for second order response surface designs using incomplete block
designs, Journal of Statistics: Advances in Theory and Applications, 10(1), 137-151.
[12] Victorbabu, B. Re. and Vasundharadevi, V. (2005). Modified second order response surface designs using balanced incomplete block designs, Sri Lankan
Journal of Applied Statistics. 6, 1-11.
[13] Victorbabu, B. Re. and Vasundharadevi, V. and Viswanadham, B. (2006). Modified second order response surface designs, rotatable designs using pairwise
balanced designs, Advances and Applications in Statistics, 6, 323-334.
[14] Victorbabu, B. Re. and Vasundharadevi, V. (2008). Modified second order response surface designs, rotatable designs using symmetrical unequal block
arrangements with two unequal block sizes, Pakistan Journal of Statistics, 24(1), 67-76.
[15] Victorbabu, B. Re. and Vasundharadevi, V. and Viswanadham, B. (2008). Modified second order response surface designs using central composite designs,
Canadian Journal of Pure and Applied Sciences, 2(1), 289-294.

©IJRASET: All Rights are Reserved 514