Chapter 5

Ratio and Product Methods of Estimation

An important objective in any statistical estimation procedure is to obtain the estimators of parameters of
interest with more precision. It is also well understood that incorporation of more information in the
estimation procedure yields better estimators, provided the information is valid and proper. Use of such
auxiliary information is made through the ratio method of estimation to obtain an improved estimator of
the population mean. In ratio method of estimation, auxiliary information on a variable is available, which
is linearly related to the variable under study and is utilized to estimate the population mean.

Let Y be the variable under study and X be an auxiliary variable which is correlated with Y . The
observations xi on X and yi on Y are obtained for each sampling unit. The population mean X of X

(or equivalently the population total X tot ) must be known. For example, xi ' s may be the values of yi ' s

from
- some earlier completed census,
- some earlier surveys,
- some characteristic on which it is easy to obtain information etc.

For example, if yi is the quantity of fruits produced in the ith plot, then xi can be the area of ith plot or the

production of fruit in the same plot in the previous year.

Let ( x1 , y1 ), ( x2 , y2 ),..., ( xn , yn ) be the random sample of size n on the paired variable (X, Y) drawn,

preferably by SRSWOR, from a population of size N. The ratio estimate of the population mean Y is

YˆR = X = RX
y ˆ
x
N
assuming the population mean X is known. The ratio estimator of population total Ytot =  Yi is
i =1

y
YˆR (tot ) = tot X tot
xtot
N n n
where X tot =  X i is the population total of X which is assumed to be known, ytot =  yi and xtot =  xi
i =1 i =1 i =1

are the sample totals of Y and X respectively. The YˆR (tot ) can be equivalently expressed as

y
YˆR (tot ) = X tot
x
ˆ .
= RX tot

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 1
 Ytot
Looking at the structure of ratio estimators, note that the ratio method estimates the relative change
X tot
yi
that occurred after ( xi , yi ) were observed. It is clear that if the variation among the values of and is
xi
ytot y
nearly same for all i = 1,2,...,n then values of (or equivalently ) vary little from sample to sample
xtot x
and the ratio estimate will be of high precision.

Bias and mean squared error of ratio estimator:

Assume that the random sample ( xi , yi ), i = 1, 2,..., n is drawn by SRSWOR and population mean X is

known. Then
N
 
n
E (YˆR ) =
1 yi

 N  i =1 xi
X
 
n
 Y (in general).

 y2 
Moreover, it is difficult to find the exact expression for E   and E  2  . So we approximate them and
y
x x 
proceed as follows:
Let
y −Y
0 =  y = (1 +  o )Y
Y
x−X
1 =  x = (1 + 1 ) X .
X
Since SRSWOR is being followed, so
E ( 0 ) = 0
E (1 ) = 0
1
E ( 02 ) = 2
E ( y − Y )2
Y
1 N −n 2
= 2 SY
Y Nn
f SY2
=
n Y2
f
= CY2
n
N −n 2 1 N S
where f =
N
, SY = 
N − 1 i =1
(Yi − Y ) 2 and CY = Y is the coefficient of variation related to Y.
Y

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 2
Similarly,
f 2
E (12 ) = CX
n
1
E ( 01 ) = E[( x − X )( y − Y )]
XY
1 N −n 1 N
=
XY Nn N − 1 i =1
 ( X i − X )(Yi − Y )
1 f
= . S XY
XY n
1 f
=  S X SY
XY n
f S S
=  X Y
n X Y
f
=  C X CY
n
SX
where C X = is the coefficient of variation related to X and  is the population correlation coefficient
X
between X and Y.

Writing YˆR in terms of  ' s, we get

YˆR = X
y
x
(1 +  0 )Y
= X
(1 + 1 ) X
= (1 +  0 )(1 + 1 ) −1Y

Assuming 1  1, the term (1 + 1 ) −1 may be expanded as an infinite series and it would be convergent.

x−X
Such an assumption means that  1, i.e., a possible estimate x of the population mean X lies
X

between 0 and 2 X . This is likely to hold if the variation in x is not large. In order to ensure that variation
in x is small, assume that the sample size n is fairly large. With this assumption,

YˆR = Y (1 +  0 )(1 − 1 + 12 − ...)

= Y (1 +  0 − 1 + 12 − 1 0 + ...).

So the estimation error of YˆR is

YˆR − Y = Y ( 0 − 1 + 12 − 1 0 + ...).

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 3
In case, when the sample size is large, then  0 and 1 are likely to be small quantities and so the terms

involving second and higher powers of  0 and 1 would be negligibly small. In such a case

YˆR − Y Y ( 0 − 1 )
and
E (YˆR − Y ) = 0.
So the ratio estimator is an unbiased estimator of the population mean up to the first order of
approximation.
If we assume that only terms of  0 and 1 involving powers more than two are negligibly small (which is

more realistic than assuming that powers more than one are negligibly small), then the estimation error of

YˆR can be approximated as

YˆR − Y Y ( 0 − 1 + 12 − 1 0 )

Then the bias of YˆR is given by

 
E (YˆR − Y ) = Y  0 − 0 + C X2 −  C X C y 
f f
 n n 
Bias(YˆR ) = E (YˆR − Y ) = YC X (C X −  CY ).
f
n
upto the second order of approximation. The bias generally decreases as the sample size grows large.

The bias of YˆR is zero, i.e.,

Bias(YˆR ) = 0
if E (12 −  01 ) = 0
Var ( x ) Cov( x , y )
or if − =0
X2 XY
1  X 
or if 2 Var ( x ) − Cov( x , y )  = 0
X  Y 
Cov( x , y )
or if Var ( x ) − = 0 (assuming X  0)
R
Y Cov( x , y )
or if R = =
X Var ( x )
which is satisfied when the regression line of Y on X passes through the origin.
Now, to find the mean squared error, consider

MSE (YˆR ) = E (YˆR − Y ) 2

= E Y 2 ( 0 − 1 + 12 − 1 0 + ...) 2 
E Y 2 ( 02 + 12 − 2 01 )  .

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 4
Under the assumption 1  1 and the terms of  0 and 1 involving powers, more than two are negligibly

small,
f 
MSE (YˆR ) = Y 2  C X2 + CY2 −
f 2f
 C X CY 
n n n 
2
Y f
= C X2 + CY2 − 2  C X C y 
n
up to the second-order of approximation.

Efficiency of ratio estimator in comparison to SRSWOR

Ratio estimator is a better estimate of Y than sample mean based on SRSWOR if

MSE (YˆR )  VarSRS ( y )

f f
or if Y 2 (C X2 + CY2 − 2  C X CY )  Y 2 CY2
n n
or if C X − 2  C X CY  0
2

1 CX
or if   .
2 CY

Thus ratio estimator is more efficient than the sample mean based on SRSWOR if
1 CX
 if R  0
2 CY
1 CX
and   − if R  0.
2 CY

It is clear from this expression that the success of ratio estimator depends on how close is the auxiliary
information to the variable under study.

Upper limit of ratio estimator:

Consider
Cov( Rˆ , x ) = E ( Rˆ x ) − E ( Rˆ ) E ( x )
y 
= E x  − E ( Rˆ ) E ( x )
x 
ˆ
= Y − E ( R) X .

Thus
Y Cov ( Rˆ , x )
E ( Rˆ ) = −
X X
Cov( Rˆ , x )
= R−
X

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 5
 Bias ( Rˆ ) = E ( Rˆ ) − R
Cov( Rˆ , x )
=−
X
 Rˆ , x  Rˆ x
=−
X

where  Rˆ , x is the correlation between Rˆ and x ;  Rˆ and  x are the standard errors of Rˆ and x

respectively.
Thus

−  Rˆ , x  Rˆ x
Bias ( Rˆ ) =
X
 Rˆ x

X
( Rˆ , x
1 . )
assuming X  0. Thus

Bias ( Rˆ ) x

 Rˆ X

Bias ( Rˆ )
or  CX
 Rˆ
where C X is the coefficient of variation of X. If C X < 0.1, then the bias in R̂ may be safely regarded as

negligible in relation to the standard error of Rˆ .

Alternative form of MSE (YˆR )

Consider
N N 2

 (Yi − RX i )2 =  (Yi − Y ) + (Y − RX i ) 
i =1 i =1
N 2

=  (Yi − Y ) − R( X i − X )  (Using Y = RX )
i =1
N N N
=  (Yi − Y ) 2 + R 2  ( X i − X ) 2 − 2 R  ( X i − X )(Yi − Y )
i =1 i =1 i =1
N
1
 (Yi − RX i )2 = SY2 + R 2 S X2 − 2RS XY .
N − 1 i =1

The MSE of YˆR has already been derived which is now expressed again as follows:

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 6
MSE (YˆR ) = Y 2 (CY2 + C X2 − 2  C X CY )
f
n
f  S2 S2 S 
= Y 2  Y2 + X2 − 2 XY 
n Y X XY 
f Y2  2 Y2 2 Y 
= S + 2 S X − 2 S XY 
2  Y
nY  X X 
= ( SY2 + R 2 S X2 − 2 RS XY )
f
n
N
f
=  (Yi − RX i )2
n( N − 1) i =1
N −n N
= 
nN ( N − 1) i =1
(Yi − RX i ) 2 .

Estimate of MSE (YˆR )

Let U i = Yi − RX i , i = 1, 2,.., N then MSE of YˆR can be expressed as

f 1 N
MSE (YˆR ) =
f

n N − 1 i =1
(U i − U ) 2 = SU2
n
1 N
where SU2 = 
N − 1 i =1
(U i − U ) 2 .

Based on this, a natural estimator of MSE (YˆR ) is

MSE (YˆR ) = su2

f
n
1 n
where su2 = 
n − 1 i =1
(ui − u ) 2
2
1 n
=  ( yi − y ) − Rˆ ( xi − x ) 
n − 1 i =1 
= s 2 + Rˆ 2 s 2 − 2 Rs
y x
ˆ ,
xy

y
Rˆ = .
x
Based on the expression
N
MSE (YˆR ) =
f
 (Yi − RX i )2 ,
n( N − 1) i =1

an estimate of MSE (YˆR ) is

n
MSE (YˆR ) =
f

n(n − 1) i =1
ˆ )2
( yi − Rxi
.
f 2 ˆ2 2 ˆ ).
= ( s y + R sx − 2 Rs xy
n

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 7
Confidence interval of ratio estimator
If the sample is large so that the normal approximation is applicable, then the 100(1-  )% confidence

intervals of Y and R are

ˆ ˆ ˆ ˆ 
 YR − Z  Var (YR ), YR + Z  Var (YR ) 
 2 2 
and
ˆ 
 R − Z  Var ( Rˆ ), Rˆ + Z  Var ( Rˆ ) 
 2 2 
respectively where Z  is the normal derivate to be chosen for a given value of confidence coefficient
2

(1 −  ).
If ( x , y ) follows a bivariate normal distribution, then ( y − Rx ) is normally distributed. If SRS is followed
for drawing the sample, then assuming R is known, the statistic
y − Rx
N −n 2
( s y + R 2 sx2 − 2 R sxy )
Nn
is approximately N(0,1).

This can also be used for finding confidence limits, see Cochran (1977, Chapter 6, page 156) for more
details.

Conditions under which the ratio estimate is optimum

The ratio estimate YˆR is the best linear unbiased estimator of Y when

(i) the relationship between yi and xi is linear passing through origin., i.e.

yi =  xi + ei ,

where ei ' s are independent with E (ei / xi ) = 0 and  is the slope parameter

(ii) this line is proportional to xi , i.e.

Var ( yi / xi ) = E (ei2 ) = Cxi

where C is constant.

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 8
 n
Proof. Consider the linear estimate of  because ˆ =  i yi where yi =  xi + ei and i ‘s are constant.
i =1

Then ˆ is unbiased if Y =  X as E ( y ) =  X + E (ei / xi ).

If n sample values of xi are kept fixed and then in repeated sampling

n
E ( ˆ ) =  i xi 
i =1
n n
and Var ( ˆ ) =  2
i Var ( yi / xi ) = C  2
x
i i
i =1 i =1

n
So E ( ˆ ) =  when  i =1
x = 1.
i i

Consider the minimization of Var ( yi / xi ) subject to the condition for being the unbiased estimator
n


i =1
x = 1 using the Lagrangian function. Thus the Lagrangian function with Lagrangian multiplier is
i i

n
 = Var ( yi / xi ) − 2 ( i xi − 1.)
i =1
n n
= C  12 xi − 2 ( i xi − 1).
i =1 i =1

Now

= 0  i xi =  xi , i = 1, 2,.., n
 i
 n
= 0   i xi = 1
 i =1
n
Using i =1
x =1
i i

n
or x
i =1
i =1

1
or  = .
nx
Thus
1
i =
nx
n

y i
y
and so ˆ = i =1
= .
nx x
Thus ˆ is not only superior to y but also the best in the class of linear and unbiased estimators.

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 9
Alternative approach:
This result can alternatively be derived as follows:
y Y
The ratio estimator Rˆ = is the best linear unbiased estimator of R = if the following two
x X
conditions hold:
(i) For fixed x, E ( y ) =  x, i.e., the line of regression of y on x is a straight line passing through
the origin.
(ii) For fixed x , Var ( x)  x, i.e., Var ( x) =  x where  is constant of proportionality.

Proof: Let y = ( y1) , y2 ,..., yn ) ' and x = ( x1 , x2 ,..., xn ) ' be two vectors of observations on

y ' s and x ' s. Hence for any fixed x ,

E ( y) =  x
Var ( y) =  =  diag( x1 , x2 ,..., xn )

where diag( x1 , x2 ,..., xn ) is the diagonal matrix with x1 , x2 ,..., xn as the diagonal elements.

The best linear unbiased estimator of  is obtained by minimizing

S 2 = ( y −  x ) '  −1 ( y −  x )
n
( yi −  xi ) 2
= .
i =1  xi
Solving
S 2
=0

n
  ( yi − ˆ xi ) = 0
i =1

y
or ˆ = = Rˆ .
x
ˆ = Yˆ is the best
Thus R̂ is the best linear unbiased estimator of R . Consequently, RX R

linear unbiased estimator of Y .

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 10
Ratio estimator in stratified sampling
Suppose a population of size N is divided into k strata. The objective is to estimate the population mean Y
using the ratio method of estimation.

In such a situation, a random sample of size ni is being drawn from the ith strata of size N i on the variable

under study Y and auxiliary variable X using SRSWOR.

Let
yij : jth observation on Y from ith strata

xij : jth observation on X from ith strata i =1, 2,…,k; j = 1, 2,..., ni .

An estimator of Y based on the philosophy of stratified sampling can be derived in the following two
possible ways:

1. Separate ratio estimator

- Employ first the ratio method of estimation separately in each stratum and obtain ratio estimator

YˆRi i = 1, 2,.., k , assuming the stratum mean X i to be known.

- Then combine all the estimates using weighted arithmetic mean.

This gives the separate ratio estimator as

k N Yˆ
ˆ
YRs = 
i Ri

i =1 N
k
=  wiYˆRi
i =1

k
yi
=  wi Xi
i =1 xi
ni
1
where yi =
ni
yj =1
ij : sample mean of Y from ith strata

ni
1
xi =
ni
x
j =1
ij : sample mean of X from ith strata

Ni
1
Xi =
Ni
x
j =1
ij : mean of all the X units in ith stratum

No assumption is made that the true ratio remains constant from stratum to stratum. It depends on
information on each X i .

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 11
2. Combined ratio estimator:
- Find first the stratum mean of Y ' s and X ' s as
k
yst =  wi yi
i =1
k
xst =  wi xi .
i =1
- Then define the combined ratio estimator as

YˆRc = st X
y
xst
N
where X is the population mean of X based on all the N =  Ni units. It does not depend on individual
i =1

stratum units. It does not depend on information on each X i but only on X .

Properties of separate ratio estimator:

k k
Note that there is an analogy between Y =  wY
i i and YRs =  wY
i Ri .
i =1 i =1

We already have derived the approximate bias of YˆR = X as

y
x

E (YˆR ) = Y +
Yf
(Cx2 −  C X CY ) .
n

So for YˆRi , we can write

E (YˆRi ) = Yi + Yi i (Cix2 − i CiX CiY )

f
ni
Ni Ni
1 1
where Yi =
Ni
 yij , X i =
j =1 Ni
x
j =1
ij

N i − ni 2 Siy
2
S2
fi = , Ciy = 2 , Cix2 = ix2 ,
Ni Yi Xi
1 Ni 1 Ni
Siy2 =  ij i ix N − 1 
N i − 1 j =1
(Y − Y ) 2
, S 2
=
j =1
( X ij − X i ) 2 ,
i

i : correlation coefficient between the observation on X and Y in ith stratum

Cix : coefficient of variation of X values in ith sample.
Thus
k
E (YˆRs ) =  wi E (YˆRi )
i =1
k
 f 
=  wi Yi + Yi i (Cix2 − i Cix Ciy ) 
i =1  ni 
k
wY f
= Y +  i i i (Cix2 − i Cix Ciy )
i =1 ni

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 12
 Bias (YˆRs ) = E (YˆRs ) − Y
k
wiYi fi
= Cix (Cix − i Ciy )
i =1 ni
upto the second order of approximation.
Assuming finite population correction to be approximately 1, ni = n / k and Cix , Ciy and i are the same

for all the strata as Cx , C y and  respectively, we have

Bias (YˆRs ) = (Cx2 −  CxC y ) .

k
n
Thus the bias is negligible when the sample size within each stratum should be sufficiently large and YRs is

unbiased when Cix =  Ciy .

Now we derive the approximate MSE of YˆRs . We already have derived the MSE of YˆR earlier as

Y2f
MSE (YˆR ) = (C X2 − CY2 − 2  C xC y )
n
N
f
= 
n( N − 1) i =1
(Yi − RX i ) 2

Y
where R = .
X
Thus the MSE of ratio estimate up to the second order of approximation based on ith stratum is

MSE (YˆRi ) =
fi
(CiX2 + CiY2 − 2 i CiX CiY )
ni ( N i − 1)
Ni
fi
= 
ni ( N i − 1) j =1
(Yij − Ri X ij ) 2

and so
k
MSE (YˆRs ) =  wi2 MSE (YˆRi )
i =1
k
 w2 f 
=   i i Yi 2 (CiX2 + CiY2 − 2 i CiX CiY ) 
i =1  ni 
k  Ni

fi
=   wi2  (Yij − Ri X ij ) 2  .
i =1  ni ( N i − 1) j =1 

An estimate of MSE (YˆRs ) can be found by substituting the unbiased estimators of SiX2 , SiY2 and SiXY
2
as

six2 , siy2 and sixy , respectively for ith stratum and Ri = Yi / X i can be estimated by ri = yi / xi .

ˆ
k
 wi2 fi 2 
MSE (YRs ) =   ( siy + ri 2 six2 − 2ri sixy ) .
i =1  ni 
Also
k  
wi2 fi ni
MSE (YˆRs ) =    ( yij − ri xij ) 2 
i =1  ni ( ni − 1) j =1 

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 13
Properties of combined ratio estimator:
Here
k

w y i i
YˆRC =
yst
i =1
k
X= X = Rˆc X .
w x
xst
i i
i =1

It is difficult to find the exact expression of bias and mean squared error of YˆRc , so we find their

approximate expressions.

Define
yst − Y
1 =
Y
x −X
 2 = st
X
E (1 ) = 0
E ( 2 ) = 0
k
N i − ni wi2 SiY2 k
f i wi2 SiY2  ˆ , E ( 2 ) = f SY = f C 2 
2
E (12 ) =  =   Recall that in case of Y R 1 Y 
i =1 N i ni Y 2 i =1 ni Y2  n Y2 n 
fi wi2 SiX2
k
E ( ) = 
2
2
i =1 ni X2
k
fi SiXY
E (1 2 ) =  wi2 .
i =1 ni XY

Thus assuming  2  1,

(1 + 1 )Y
YˆRC = X
(1 +  2 ) X
= Y (1 + 1 )(1 −  2 +  22 − ...)
= Y (1 + 1 −  2 − 1 2 +  22 − ...)

Retaining the terms up to order two due to the same reason as in the case of YˆR ,

YˆRC Y (1 + 1 −  2 − 1 2 +  22 )
Yˆ − Y = Y ( −  −   +  2 )
RC 1 2 1 2 2

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 14
The approximate bias of YˆRc up to the second-order of approximation is

Bias (YˆRc ) = E (YˆRc − Y )

YE (1 −  2 − 1 2 +  22 )
= Y 0 − 0 − E ( 1 2 ) + E (  22 ) 
k 
f  S 2 S 
= Y   i wi2  iX2 − iXY  
i =1  ni X XY  
k 
f  S 2  S S 
= Y   i wi2  iX2 − i iX iY  
i =1  ni X XY  
Y k  fi S  S 
= n wi2 SiX  iX − i iY  
X i =1  i  X Y 
k
f 
= R   i wi2 SiX ( CiX − i CiY ) 
i =1  ni 
Y
where R = , i is the correlation coefficient between the observations on Y and X in the ith stratum,
X
Cix and Ciy are the coefficients of variation of X and Y respectively in the ith stratum.

The mean squared error upto second order of approximation is

MSE (YˆRc ) = E (YˆRc − Y ) 2

Y 2 E (1 −  2 − 1 2 +  2 ) 2
Y 2 E (12 +  22 − 21 2 )
k  f i 2  SiX2 SiY2 2SiXY  
=Y   wi  2 + 2 −
2


i =1  ni X Y XY  
k 
f  S 2 S 2 2 S S 
= Y 2   i wi2  iX2 + iY2 − i iX iY  
i =1  ni X Y X Y 
Y2 k  fi 2  Y 2 2 Y 
=
Y2
  wi  2 SiX + SiY − 2 i SiX SiY  
2

i =1  ni X X 
k
f 
=   i wi2 ( R 2 SiX2 + SiY2 − 2 i RSiX SiY ) .
i =1  ni 

An estimate of MSE (YRc ) can be obtained by replacing SiX2 , SiY2 and SiXY by their unbiased estimators

Y y
six2 , siy2 and sixy respectively whereas R = is replaced by r = . Thus the following estimate is
X x
obtained:
 w2 f 
MSE (YRc ) =   i i ( r 2 six2 + siy2 − 2rsixy ) .
k

i =1  ni 
Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 15
Comparison of combined and separate ratio estimators
An obvious question arises that which of the estimates YˆRs or YˆRc is better. So we compare their MSEs.

Note that the only difference in the term of these MSEs is due to the form of ratio estimate. It is

in MSE (YˆRs )
yi
− Ri =
xi

in MSE (YˆRc ).
Y
− R=
X
Thus

 = MSE (YˆRc ) − MSE (YˆRs )

k
 w2 f 
=   i i ( R 2 − Ri2 ) SiX2 + 2( Ri − R) i SiX SiY  
i =1  ni 
k
 w2 f 
=   i i ( R − Ri ) 2 SiX2 + 2( R − Ri )( Ri SiX2 − i SiX SiY )  .
i =1  ni 

The difference  depends on

(i) The magnitude of the difference between the strata ratios ( Ri ) and whole population ratio (R).

(ii) The value of ( Ri Six2 − i Six Siy ) is usually small and vanishes when the regression line of y on x is

linear and passes through origin within each stratum. See as follows:
Ri Six2 − i Six Siy = 0
i Six Siy
Ri =
Six2
which is the estimator of the slope parameter in the regression of y on x in the ith stratum. In
such a case

MSE (YˆRc )  MSE (YˆRs )

but Bias (Yˆ )  Bias(Yˆ ).
Rc Rs

So unless Ri varies considerably, the use of YˆRc would provide an estimate of Y with negligible bias and

precision as good as YˆRs .

• If Ri  R, YˆRs can be more precise but bias may be large.

If Ri R, YˆRc can be as precise as YˆRs but its bias will be small. It also does not require knowledge
•
of X 1 , X 2 ,..., X k .

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 16
Ratio estimators with reduced bias:
The ratio type estimators that are unbiased or have smaller bias than Rˆ , YˆR or YˆRc (tot ) are useful in sample

surveys. There are several approaches to derive such estimators. We consider here two such approaches:

1. Unbiased ratio – type estimators:

Y
Under SRS, the ratio estimator has form X to estimate the population mean Y . As an alternative to
x
this, we consider following as an estimator of the population mean
1 n Y 
YˆRo =   i  X .
n i =1  X i 
Yi
Let Ri = , i = 1, 2,.., N ,
Xi
then

1 n
YˆR 0 =  Ri X
n i =1
= rX
where
1 n
r =  Ri
n i =1
Bias (YˆR 0 ) = E (YˆR 0 ) − Y
= E (rX ) − Y
= E (r ) X − Y .
Since
1 n 1 N
E (r ) = (
n i =1 N
R )
i =1
i

1 n
= R
n i =1
= R.
So Bias (YˆR 0 ) = RX − Y .

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 17
 N −n
Using the result that under SRSWOR, Cov( x , y ) = S XY , it also follows that
Nn
N −n 1 N
Cov(r , x ) =  ( Ri − R )( X i − X )
Nn N − 1 i =1
N −n 1 N
= ( Ri X i − NRX )
Nn N − 1 i =1
N −n 1 N
Y
= ( i X i − NRX )
n N − 1 i =1 X i
N −n 1
= ( NY − NRX )
Nn N − 1
N −n 1
= [− Bias (YˆR 0 )].
n N −1
N −n N −n
Thus using the result that in SRSWOR, Cov( x , y ) = S XY , and therefore Cov(r , x ) = S RX , we
Nn Nn
have
n( N − 1)
Bias (YˆRo ) = − Cov(r , x )
N −n
n( N − 1) N − n
=− S RX
N − n Nn
 N −1 
= −  S RX
 N 
1 N
where S RX =  ( Ri − R )( X i − X ).
N − 1 i =1
The following result helps in obtaining an unbiased estimator of a population mean:
Since under SRSWOR set up,
E ( sxy ) = S xy
1 n
where sxy =  ( xi − x )( yi − y ),
n − 1 i =1
1 N
S xy =  ( X i − X )(Yi − Y ).
N − 1 i =1
So an unbiased estimator of the bias in Bias(YˆR 0 ) = −( N − 1) S RX is obtained as follows:

( N − 1)
Bias (YˆR 0 ) = − srx
N
N −1 n
=−  (ri − r )( xi − x )
N (n − 1) i =1
N −1 n
=− ( ri xi − n r x )
N (n − 1) i =1
N − 1  n yi 
=−   xi − nr x 
N (n − 1)  i =1 xi 
N −1
=− (ny − nr x ).
N (n − 1)

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 18
So

( )
Bias (YˆR 0 ) = E YˆR 0 − Y = −
n( N − 1)
N (n − 1)
( y − r x ).

Thus
E YˆR 0 − Bias (YˆR 0 )  = Y
 
 n( N − 1) 
or E YˆR 0 + ( y − r x ) = Y .
 N (n − 1) 
Thus
n( N − 1) n( N − 1)
YˆR 0 + ( y − r x ) = rX + (y − r x)
N (n − 1) N (n − 1)
is an unbiased estimator of the population mean.

It may noted that such an unbiased estimator cannot be obtained using YˆR = X because it is not exactly
y
x

unbiased. It is unbiased only up to the first order of approximation. So even if the bias of YˆR = X upto
y
x
the first order of approximation is used to obtain such an unbiased estimator, the estimator will change for
higher order of approximations.

2. Jackknife method for obtaining a ratio estimate with lower bias

Jackknife method is used to get rid of the term of order 1/n from the bias of an estimator. Suppose the
E ( Rˆ ) can be expanded after ignoring finite population correction as
a a
E ( Rˆ ) = R + 1 + 22 + ...
n n

Let n = mg and the sample is divided at random into g groups, each of size m. Then
ga1 ga
E ( gRˆ ) = gR + + 2 2 2 + ...
gm g m
a a
= gR + 1 + 2 2 + ...
m gm

Let Rˆi* =  * i where the

*
y
 *
denotes the summation over all values of the sample except the ith group.
 xi
So Rˆi* is based on a simple random sample of size m(g - 1), so we can express

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 19
 a1 a
E ( Rˆi* ) = R + + 2 2 2 + ...
m( g − 1) m ( g − 1)
or
a a
E ( g − 1) Rˆi*  = ( g − 1) R + 1 + 2 2 + ...
m m ( g − 1)
Thus
a2
E  gRˆ − ( g − 1) Rˆi*  = R − + ...
g ( g − 1)m 2
or
a2 g
E  gRˆ − ( g − 1) Rˆi*  = R − + ...
n2 g − 1
1
Hence the bias of  gRˆ − ( g − 1) Rˆi*  is of order 2 .
n
Now g estimates of this form can be obtained, one estimator for each group. Then the jackknife or
Quenouille’s estimator is the average of these of estimators
g

 Rˆ i
RˆQ = gRˆ − ( g − 1) i =1
.
g

Product method of estimation:

1 C
The ratio estimator is more efficient than the sample mean under SRSWOR if   . x , if R  0, which
2 Cy

1 Cx
is usually the case. This shows that if auxiliary information is such that   − , then we cannot use
2 Cy

the ratio method of estimation to improve the sample mean as an estimator of the population mean. So
there is a need for another type of estimator which also makes use of information on auxiliary variable X.
Product estimator is an attempt in this direction.
The product estimator of the population mean Y is defined as

YˆP =
yx
.
X
assuming the population mean X to be known

We now derive the bias and variance of Yˆp .

y −Y x−X
Let  0 = , 1 = ,
Y X

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 20
(i) Bias of Yˆp .

We write Yˆp as

Yˆp =
yx
= Y (1 +  0 )(1 + 1 )
X
= Y (1 +  0 + 1 +  01 ).

Taking expectation, we obtain bias of Yˆp as

Bias(Yˆp ) = E ( 01 ) Cov( y , x ) =

1 f
S xy ,
X nX

which shows that bias of Yˆp decreases as n increases. Bias of Yˆp can be estimated by

Bias(Yˆp ) =
f
sxy .
nX

(ii) MSE of Yˆp :

Writing Yˆp is terms of  0 and 1 , we find that the mean squared error of the product estimator Yˆp up to

second order of approximation is given by

MSE (Yˆp ) = E (Yˆp − Y ) 2

= Y 2 E (1 +  0 +  01 )2
 Y 2 E (12 +  02 + 2 01 ).

Here terms in (1 ,  0 ) of degrees greater than two are assumed to be negligible. Using the expected values,

we find that

MSE (Yˆp ) =  SY2 + R 2 S X2 + 2 RS XY  .

f
n

(iii) Estimation of MSE of Yˆp

The mean squared error of Yˆp can be estimated by

MSE (Yˆp ) =  s y2 + r 2 sx2 + 2rsxy 

f
n
where r = y / x .

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 21
(iv) Comparison with SRSWOR:
From the variances of the sample mean under SRSWOR and the product estimator, we obtain

Var ( y ) SRS − MSE (Yˆp ) = − RS X (2  SY + RS X ),

f
n
f 2
where Var ( y ) SRS = SY which shows that Yˆp is more efficient than the simple mean y for
n
1 Cx
− if R  0
2 Cy

and for
1 Cx
 − if R  0.
2 Cy

Multivariate Ratio Estimator

Let y be the study variable and X 1 , X 2 ,..., X p be p auxiliary variables assumed to be corrected with y .

Further, it is assumed that X 1 , X 2 ,..., X p are independent. Let Y , X 1 , X 2 ,..., X p be the population means of

the variables y , X 1 , X 2 ,..., X p . We assume that a SRSWOR of size n is selected from the population of

N units. The following notations will be used.

Si2 : the population mean sum of squares for the variate X i ,
si2 : the sample mean sum of squares for the variate X i ,
S02 : the population mean sum of squares for the study variable y,
s02 : the sample mean sum of squares for the study variable y,
Si
Ci = : coefficient of variation of the variate X i ,
Xi
S0
C0 = : coefficient of variation of the variate y,
Y
S
i = iy : coefficient of correlation between y and X i ,
Si S 0

YˆRi =
y
:ratio estimator of Y , based on X i
Xi

where i = 1, 2,..., p. Then the multivariate ratio estimator of Y is given as follows.

p p
YˆMR =  wiYˆRi , w =1 i
i =1 i =1
p
Xi
= y  wi .
i =1 xi

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 22
(i) Bias of the multivariate ratio estimator:
The approximate bias of YˆRi up to the second order of approximation is

Bias(YˆRi ) = Y (Ci2 − iCiC0 ).

f
n

The bias of YˆMR is obtained as

p
Bias (YˆMR ) =  wi
Yf
(Ci2 − i Ci C0 )
i =1 n
p
Yf
=
n
 w C (C
i =1
i i i − i C0 ).

(ii) Variance of the multivariate ratio estimator:

The variance of YˆRi up to the second-order of approximation is given by

Var (YˆRi ) = Y 2 (C02 + Ci2 − 2 iC0Ci ).

f
n

The variance of YˆMR up to the second-order of approximation is obtained as

p
Var (YˆMR ) = Y 2  wi2 (C02 + Ci2 − 2 iC0Ci ).
f
n i =1

Sampling Theory| Chapter 5 | Ratio & Product Methods of Estimation | Shalabh, IIT Kanpur
Page 23