CHAPTER FIVE

RATIO ESTIMATORS
5.1. Estimation of a Ratio under Simple Random Sampling
We often have auxiliary information in surveys; few investigators go to the expense of taking a
good sample and then measure only one quantity. Often the sampling frame gives us extra
information about each unit that can be used to improve the precision of our estimates.
In the ratio method an auxiliary variate Xi correlated with Yi is obtained for each unit in the same
sample. The population total X of the Xi must be known. In practice Xi is often the value of Yi at
some previous time when a complete census was taken. The aim in this method is to obtain
increased precision by taking advantage of the correlation between Yi and Xi.
Suppose the units of the population possess two characteristics that are correlated to each other.
Let two varieties Xi and Yi, i=1, 2, …, N, represent the two characteristics. Assume that Y
represents the study variable and X represents an auxiliary variable. The ratio of these two
variables is the simplest and most commonly used method of the complex estimation techniques
for improving the precision or reliability.
For ratio estimator we select a SRS of size n out of a population of size N and measure both the
variables Yi (the variable under study) and Xi (the auxiliary variable) on the units in the sample.
Let the values in the sample be denoted by
(𝑦𝑖 , 𝑥𝑖 ), 𝑖 = 1,2, … , 𝑛
We assume that the population mean (𝑋̅)of the auxiliary variable is known and the interest is to
estimate the ratio(R), the mean and the total of study variable. In the population of size we have
∑𝑁 𝑌 𝑌 𝑌̅
𝑌 = ∑𝑁 𝑁 𝑖=1 𝑖
𝑖=1 𝑌𝑖 And 𝑋 = ∑𝑖=1 𝑋𝑖 , the ratio can be defined as: 𝑅 = ∑𝑁 𝑋 = 𝑋 = 𝑋̅
𝑖=1 𝑖

Define the population correlation coefficient between X and Y to be

𝑆𝑥𝑦 (𝑋𝑖 −𝑋̅ )(𝑌𝑖 −𝑌̅)
𝜌=𝑆 , where 𝑆𝑥𝑦 = ∑𝑁
𝑖=1 is the covariance between X and Y and 𝑆𝑥 and 𝑆𝑦 are
𝑥 𝑆𝑦 𝑁−1
standard derivations of X and Y respectively.
The ratio estimator of the population mean 𝑌̅ is defined by
𝑦̅ 𝑦̅
𝑌̅̂𝑅 = 𝑋̅ = 𝑅̂ 𝑋̅ , Where 𝑋̅ is known 𝑅̂ = .
𝑥̅ 𝑥̅
For each unit in the sample of size n elements selected by SRS, there are sample values:
∑ 𝑛
𝑦𝑖 𝑦 𝑦̅
y={y1,y2,…,yn} and x={x1,x2,…,xn}. The sample estimate for the ratio is 𝑅̂ =∑𝑖=1
𝑛 𝑥𝑖 = 𝑥 = 𝑥̅ .
𝑖=1

The estimate ratio,𝑅̂ , is not necessarily an unbiased estimate of a population ratio, R, that is
𝐸(𝑅̂ ) ≠ 𝑅. In most situations the bias is small, and estimated ratios are widely used. For large
samples, mostly if n≥30, the ratio estimate is a consistent estimate, which tends to normality with

Sampling Theory | Chapter FIVE | Ratio Estimators 1

 𝑦̅
negligible bias. For ratio estimate, 𝑅̂ = both the numerator (𝑦̅) and the denominator (𝑥̅ ) are
𝑥̅
subject to sampling variability. Therefore, an exact expression for its standard error is complicated,
but in large samples its approximation is valid.
Consider the following examples in which the variables are somehow related:
 Percentage of food expenditure to that of total income;
 The use of fertilizer per capital;
 Gross enrolment ratio;
Theorem 5.1: If variates yi and xi are measured on each unit of a simple random sample of size n,
𝑦̅
assumed large, the MSE and variance of 𝑅̂ = are each approximately 𝑀𝑆𝐸(𝑅̂ ) = 𝑉(𝑅̂ ) =
𝑥̅
1−𝑓 ∑𝑁𝑖=1(𝑦𝑖−𝑅𝑥𝑖)
2 𝑌̅ 𝑛
; where 𝑅 = is the ratio of the population means and 𝑓 = 𝑁 If we have a sample
𝑛𝑋̅ 2 𝑁−1 𝑋̅
estimate 𝑅̂ from n sample observations, then the variance of the estimated ratio is given by 𝑉(𝑅̂ ) =
1−𝑓 ∑𝑛 ̂ 2
𝑖=1(𝑦𝑖−𝑅 𝑥𝑖) 1−𝑓 ∑𝑛 ̂ 2
𝑖=1(𝑦𝑖−𝑅 𝑥𝑖)
= 𝑛𝑋̅ 2 𝑆𝑑2 , where 𝑆𝑑2 = .If 𝑋̅ is unknown, substitute it by its sample
𝑛𝑋̅ 2 𝑛−1 𝑛−1
estimate, 𝑥̅ .
𝑛 ̂ 2
1−𝑓 ∑ (𝑦𝑖−𝑅 𝑥𝑖) 1−𝑓 √1−𝑓 𝑆
The standard error of 𝑅̂ is 𝑠. 𝑒. (𝑅̂ ) = √𝑛𝑋̅ 2 𝑖=1 𝑛−1 = √𝑛𝑋̅ 2 𝑆𝑑2 = 𝑛 𝑋̅𝑑
√

5.1. The ratio Estimator and Its Variance

Assuming simple random sampling and considering the advantage of the correlation between Xi
and Yi, we can estimate the population total of the Yi, (Y) and the mean (𝑌̅) using the ratio estimate
as stated the following theorem.
Theorem 5.2: The ratio estimate of the population total Y and the population mean 𝑌̅are,
respectively,
𝑦̅ 𝑦̅
𝑌̂𝑅 = 𝑋, 𝑌̅̂𝑅 = 𝑋̅, assuming X is known. In simple random sample of size n, with large n, the
𝑥̅ 𝑥̅
variances of ratio estimates of population total and mean are given by:
𝑁 2
1−𝑓 ∑ (𝑦−𝑅𝑥𝑖) 1−𝑓
𝑉(𝑌̂𝑅 ) = 𝑁 2 𝑛 𝑖=1 𝑁−1 = 𝑁2 𝑛
𝑆𝑑2 ,
𝑁 2
1−𝑓 ∑ (𝑦−𝑅𝑥𝑖) 1−𝑓
𝑉(𝑌̅̂𝑅 ) = 𝑛 𝑖=1 𝑁−1 = 𝑆𝑑2
𝑛
Corollary:
1−𝑓
i) 𝑉(𝑌̂𝑅 ) = 𝑁 2 𝑛 (𝑆𝑦2 + 𝑅 2 𝑆𝑥2 − 2𝑅 𝑆𝑦 𝑆𝑋 )
𝑆𝑦 2
2𝑆𝑥𝑦 2
1−𝑓 𝑆
ii) 𝑉(𝑌̂𝑅 ) = 𝑛 𝑌 2 (𝑌̅ 2 + 𝑋̅𝑥2 − 𝑌̅𝑋̅ )
1−𝑓
iii) 𝑉(𝑌̂𝑅 ) = 𝑛 𝑌 2 (𝑐𝑣𝑦2 + 𝑐𝑣𝑥2 − 2𝑐𝑣𝑥𝑦 ), Where 𝑐𝑣𝑦 and 𝑐𝑣𝑥 are coefficients of variation of
yi and xi respectively, and CVxy is the relative covariance.
iv) The coefficient of variation of 𝑌̂𝑅 ,𝑌̅̂𝑅 and 𝑅̂ are the same for the three estimates, i.e.
√1−𝑓 𝑆
𝐶𝑉(𝑌̂𝑅 ) = 𝐶𝑉(𝑌̅̂𝑅 ) = 𝐶𝑉(𝑅̂ ) = 𝑛 𝑅̂𝑑𝑋̅
√

Sampling Theory | Chapter FIVE | Ratio Estimators 2

5.2. Estimation of the Variance from a Sample
In the above expressions the population parameters 𝑆𝑦2 , 𝑆𝑥2 , 𝑆𝑥𝑦 , and R are usually unknown.
Therefore, we substitute these parameters by 𝑠𝑦2 , 𝑠𝑥2 , 𝑠𝑥𝑦 , and 𝑅̂ , respectively to get the following
variances of the estimates.
𝑛 ̂ 2
1−𝑓 ∑ (𝑦𝑖−R𝑥𝑖) 1−𝑓 2 1−𝑓
i) 𝑉(𝑌̂𝑅 ) = 𝑁 2 𝑛 𝑖=1 𝑛−1 = 𝑁2 𝑠𝑑 = 𝑁2 (𝑠𝑦2 + 𝑅̂ 2 𝑠𝑥2 − 2𝑅̂ 𝑠𝑥𝑦 )
𝑛 𝑛
𝑛 2
1−𝑓 ∑ (𝑦𝑖−𝑅̂ 𝑥𝑖) 1−𝑓
ii) 𝑉(𝑌̅̂𝑅 ) = 𝑛 𝑖=1 𝑛−1 = 𝑛 (𝑠𝑦2 + 𝑅̂ 2 𝑠𝑥2 − 2𝑅̂ 𝑠𝑥𝑦 )
𝑛 ̂ 2
1−𝑓 ∑ (𝑦𝑖−𝑅 𝑥𝑖) 1−𝑓
iii) 𝑉(𝑅̂ ) = 𝑛𝑋̅ 2 𝑖=1 𝑛−1 = 𝑛𝑋̅ 2 (𝑠𝑦2 + 𝑅̂ 2 𝑠𝑥2 − 2𝑅̂ 𝑠𝑥𝑦 ) If 𝑋̅ is unknown, we substitute
by the sample mean𝑥̅ .
Confidence Limits
A 100(1-ɑ) % confidence interval may be constructed for large samples to apply normal
distribution.
 For total, 𝑌: 𝑌̂𝑅 ± 𝑍ɑ 𝑠. 𝑒(𝑌̂𝑅 ),
2

 For mean, 𝑌̅: 𝑌̅̂𝑅 ± 𝑍ɑ 𝑠. 𝑒(𝑌̅̂𝑅 )

2
 For ratio, 𝑅: 𝑅̂ ± 𝑍ɑ 𝑠. 𝑒(𝑅̂ )
2
Example 1: A simple random sample of size n=33 low-income families was considered. Let x1i
represent the size of the family, yi denotes the weekly expenditure on food per family, and x2i
represent weekly family income. The following data were available on these variables.
∑𝑛𝑖=1 𝑥1𝑖 =123∑𝑛𝑖=1 𝑥2𝑖 =2394∑𝑛𝑖=1 𝑦𝑖 =907.2 ∑𝑛𝑖=1 𝑥1𝑖 𝑦𝑖 =3595.5
∑𝑛𝑖=1 𝑥1𝑖
2
=533 ∑𝑛𝑖=1 𝑥2𝑖
2
=177254 ∑𝑛𝑖=1 𝑦𝑖2 =28224 ∑𝑛𝑖=1 𝑥2𝑖 𝑦𝑖 =66678
Estimate:
a. The mean weekly expenditure on food per family.
b. The mean weekly expenditure on food per person.
c. The percentage of the income that is spent on food.
In each case compute the standard error of the estimate and ignore the fpc.
∑𝑛
𝑖=1 𝑦𝑖 907.2
Solution: a) Mean weekly expenditure on food per family:𝑦̅ = = = $27.49
𝑛 33
2
∑𝑛 2 (∑ 𝑦𝑖 ) (907.2)2⁄
1−𝑓 𝑆𝑦2 𝑖=1 𝑦𝑖 − 28224−
𝑉(𝑦̅) = 𝑆𝑦2 = = 𝑛
= 33
= 3.11 𝑠. 𝑒(𝑦̅) = 1.76
𝑛 𝑛 𝑛(𝑛−1) 33(32)
b) Let 𝑅̂1 = mean weekly expenditure on food per person.
∑ 𝑛
𝑦𝑖 907.2 ∑𝑛
𝑖=1 𝑥1 907.2
𝑅̂ 1=∑𝑛𝑖=1𝑥1𝑖= 123 =$7.38, the average percent per family is 𝑥̅ 1= 𝑖
= =3.7273.
𝑖=1 𝑛 33
Then the variance of𝑅̂ 1 is given as:
2 𝑛
∑ (𝑦𝑖 −𝑅1 𝑥1𝑖 ) ̂ 2 2 ̂2 2 ̂
1−𝑓 𝑠 ∑ 𝑦 + 𝑅 ∑ 𝑥1𝑖 −2𝑅1 ∑𝑖 𝑦𝑖 𝑥1𝑖
𝑉(𝑅̂1 ) = 𝑛𝑥̅ 2 𝑠𝑑 2 = 𝑛𝑥̅𝑑 2 = 𝑖=1 = 𝑖 𝑖 1𝑥̅ 2 𝑛(𝑛−1)
1 1 𝑥̅ 2 𝑛(𝑛−1) 1 1
28224+(7.38)2 (533)−2(7.38)(3595.5)
𝑉(𝑅̂1 ) = = 0.285 s.e ( 𝑅̂1 )=0.534
(3.7273)2 𝑥 33𝑥32
C) Let 𝑅̂ 2 is percentage of income spent on food.
∑ 𝑛 ∑𝑛
𝑦𝑖 907.2 𝑖=1 𝑥2𝑖 2394
𝑅̂ 2=∑𝑛𝑖=1𝑥2𝑖= 2394 𝑥100%=37.9%, 𝑥̅ 2= = = 2.5455
𝑖=1 𝑛 33

Sampling Theory | Chapter FIVE | Ratio Estimators 3

 1−𝑓 𝑠 2 ∑𝑛 ̂
𝑖=1(𝑦𝑖−𝑅2 𝑥2𝑖 )
2 ∑𝑖 𝑦𝑖2 + 𝑅̂ 22 ∑ 𝑥22 −2𝑅̂2 ∑𝑖 𝑦𝑖 𝑥2
(𝑉(𝑅̂2 ) = 𝑛𝑥̅ 2 𝑠𝑑 2 = 𝑛𝑥̅𝑑 2 = = 𝑖 𝑖
2 𝑥̅ 2 2 𝑛(𝑛−1)
2 𝑥̅ 2 2 𝑛(𝑛−1)
2
28224+(0.379) (177254)−2(0.379)(66678)
𝑉(𝑅̂2 ) = = 0.00056554 𝑠. 𝑒. (𝑅̂2 ) = 0.02378
(72.5455)2 𝑥 33𝑥32
Exercise 1: The following table shows the total hectares in 3 farms along with the payments for
farm labor drawn from 30 farms. The true value of the total hectares of all farms, X is assumed to
be 800.
Farm (I) Hectares(𝑥𝑖 ) Payments (𝑦𝑖 )
1 6 84
2 8 117
3 10 201
∑𝑛𝑖=1 𝑦𝑖 = 402, ∑𝑛𝑖=1 𝑥𝑖 = 24,, ∑𝑛𝑖=1 𝑥𝑖 2 = 200, ∑𝑛𝑖=1 𝑦𝑖 2 = 61146, ∑𝑛𝑖=1 𝑦𝑖 𝑥𝑖 = 3450
a. Estimate the total payments 𝑌̂𝑅 .
b. Estimate the variance of 𝑌̂𝑅 and its standard error.
c. Find a 95% confidence interval for Y use normal distribution with critical value= 1.96.
Example 2: A company wants to estimate the average amount of money paid to employees for
medical expenses during the three months of the current calendar year. A random sample of 100
employees’ records is taken from the population of 1000 employees and the medical expenses in the
first three months of the last year (xi) and this year (yi) recorded. From last year’s balance sheet it
was estimated that the total medical expenses in the first three months of last year were $12,500.
From the sample of employees the following result was found.
∑100 100 2 100
𝑖=1 𝑥𝑖 =1200, ∑𝑖=1 𝑥𝑖 =15,620, ∑𝑖=1 𝑦𝑖 =1750 ∑100 2 100
𝑖=1 𝑦𝑖 =31,650 ∑𝑖=1 𝑥𝑖 𝑦𝑖 =22059.35
a) Find the point estimate for the ratio.
b) Find the estimated mean expenditure for medical expenses per employee.
c) Find the standard error of both estimates.
d) Find a 95% confidence interval for the population mean.
∑𝑛 𝑦𝑖 1750 𝑦̅ 12500
Solution: a) 𝑅̂ = 𝑖=1 = ∑𝑛
=1.4583, b) 𝑌̅̂ = 𝑋̅ = 𝑅̂ 𝑋̅=1.4583(
𝑅 ) =18.23
𝑖=1 𝑥𝑖 1200 𝑥̅ 1000
c) To find standard error of the estimate, the following sample statistics must be calculated.
2 2
∑𝑛 2 (∑ 𝑥𝑖 ) (1200)2⁄ ∑𝑛 2 (∑ 𝑦𝑖 ) (1750)2⁄
𝑖=1 𝑥𝑖 − 15620− 100 𝑖=1 𝑦𝑖 − 31650− 100
𝑠𝑥2 = 𝑛
= = 12.32 , 𝑠𝑦2 = 𝑛
= = 10.35
𝑛−1 100−1 𝑛−1 100−1
(∑𝑛𝑖 𝑥𝑖 )( ∑𝑛𝑖 𝑦𝑖)⁄ (1200)(1750)
∑𝑛𝑖 𝑥𝑖 𝑦𝑖 − ( 𝑛) 22059.35 −
𝑠𝑥𝑦 = = 100 = 10.7005
𝑛−1 100 − 1
1−𝑓 1−0.1
𝑉(𝑅 ) = ̅ 2 (𝑠𝑦 + 𝑅 𝑠𝑥 − 2𝑅̂ 𝑠𝑥𝑦 ) =
̂ 2 ̂ 2 2
2 (10.35 +
(1.4583)2 (12.32) −
𝑛𝑋 100(12.5)
2(1.4583)(10.7005)) = 0.00033382 𝑠. 𝑒(𝑅̂ ) = 0.01827
1−𝑓 2
𝑉(𝑌̅̂𝑅 ) = (𝑠𝑦 + 𝑅̂ 2 𝑠𝑥2 − 2𝑅̂ 𝑠𝑥𝑦 )
𝑛
1 − 0.1
= (10.35 + (1.4583)2 (12.32) − 2(1.4583)(10.7005)) = 0.04807
100
(𝑠. 𝑒. (𝑌̅̂𝑅 ) = 0.21925

Sampling Theory | Chapter FIVE | Ratio Estimators 4

 d) 95%confidence limits for the population mean
𝑌̅=𝑌̅̂ ± 𝑍ɑ s.e (𝑌̅̂ ) = 18.23 ± 1.96 ∗ 0.21925 = (17.8,18.7)
𝑅 𝑅
2
5.3. Bias of the Ratio Estimate
There are different methods of examining the bias, but here we treat only one method which gives
an exact result for the bias and an upper bound to the ratio of the bias to the standard error. Consider
the relationship between 𝑅̂ and 𝑥̅ . That is, to consider the covariance and correlation of 𝑅̂ and 𝑥̅ in
simple random sample of size n,
Cov(𝑅̂ , 𝑥̅ )=E{[𝑅̂ − 𝐸(𝑅̂ )[𝑥̅ − 𝐸(𝑥̅ )]}=-𝑋̅𝐸(𝑅̂ − 𝑅)=− 𝑋̅𝐵𝑅̂
𝐶𝑜𝑣(𝑅̂ ,𝑥̅ ) −𝑋̅ 𝐵𝑅
Then the correlation between 𝑅̂ and 𝑥̅ is given by: 𝑅̂𝑥̅ = = ̂
𝑆𝑅
̂ 𝑆𝑥
̅ 𝑆𝑅
̂ 𝑆𝑥
̅
|𝐵𝑅
̂|
|𝐵𝑅̂ | = |𝑅̂𝑥̅ 𝑆𝑅̂ 𝑐𝑣(𝑥̅ )| ≤ 𝑐𝑣(𝑥̅ ), since |𝑅̂𝑥̅ | ≤ 1.
𝑆𝑅
̂

It shows that the magnitude of the bias in 𝑅̂ as a ratio of its standard error cannot exceed the
coefficient of variation of 𝑥̅ . If 𝑅̂ and 𝑥̅ are uncorrelated, the bias vanishes. If CV (𝑥̅ ) <10%, the bias
can be ignored. The bias in 𝑌̅̂𝑅 and 𝑌̂𝑅 can be obtained in similar way.
5.4. Comparison of Ratio estimate With the Mean per Unit
Theorem 5.3: In large samples, with simple random sample, the ratio estimate 𝑌̂𝑅 has a smaller
variance than the estimate 𝑌̂=N𝑦̅ obtained by simple expansion, if
𝑆𝑥
1( ⁄ ̅) 1 𝐶𝑉(𝑋)
𝜌> 𝑋
= 2 𝐶𝑉(𝑌), where  is the correlation between 𝑌𝑖 and 𝑋𝑖 .
2 (𝑆𝑦⁄ )
𝑌̅
𝑁 2 (1−𝑓) 1−𝑓
Proof: V(𝑌̂) = 𝑆𝑦2 , V(𝑌̂𝑅 )= N2 𝑛 (𝑠𝑦2 + 𝑅 2 𝑠𝑥2 − 2𝑅𝑆𝑥 𝑠𝑥 )
𝑛
If the V(𝑌̂𝑅 )< V(𝑌̂), then 𝑉(𝑌̂) − 𝑉(𝑌̂𝑅 ) > 0
𝑁 2 (1−𝑓) 1−𝑓
 𝑆𝑦2 - N2 (𝑠𝑦2 + 𝑅 2 𝑠𝑥2 − 2𝑅𝑆𝑦 𝑠𝑥 ) > 0
𝑛 𝑛
1 𝑅 2 𝑠𝑥2 𝑌̅
𝑆𝑦2 − 𝑆𝑦2 − 𝑅 2 𝑠𝑥2 + 2𝑅𝑆𝑦 𝑠𝑥 > 0 𝜌 > 2 𝑅𝑆 , assuming 𝑅 = 𝑋̅ is positive.
𝑦 𝑆𝑥
𝑆𝑥
1 𝑅𝑆𝑥 1( ⁄ ̅) 1 𝐶𝑉(𝑋)
>2 >2 𝑆𝑦
𝑋
= 2 𝐶𝑉(𝑌)
𝑆𝑦 ( ⁄̅)
𝑌
Therefore, if the difference between the variances of simple and ratio estimate is greater than zero,
i.e; 𝑉(𝑌̂) − 𝑉(𝑌̂𝑅 ) > 0, then a ratio estimate is more efficient. If the difference is zero, then both
estimates are equally efficient. If the difference is less than zero, the ratio estimate is not as efficient
as the estimate from simple expansion.
5.5. Ratio Estimate in Stratified Random Sampling
In stratified random sampling design, there are two methods for estimating ratios: the separate ratio
estimate and the combined ratio estimate.
𝑦 𝑦̅
The separated ratio estimate: For stratum h:𝑌̂𝑅ℎ = ℎ 𝑋ℎ = ℎ 𝑋ℎ =𝑅̂ℎ 𝑋ℎ , and its variance will be
𝑥ℎ 𝑥̅ ℎ
𝑵𝟐𝒉 (𝟏−𝒇𝒉 )
𝑉(𝑌̂𝑅ℎ ) = 2
(𝑠𝑦ℎ + 𝑅ℎ2 𝑠𝑥ℎ
2
− 2𝑅ℎ ℎ 𝑆𝑦ℎ 𝑠𝑥ℎ ), where 𝑦ℎ 𝑎𝑛𝑑 𝑥ℎ are sample totals in the hth
𝒏𝒉
stratum, and 𝑋ℎ is the population stratum total and should be known. For the overall total, the
Sampling Theory | Chapter FIVE | Ratio Estimators 5
 𝑦 𝑦̅
separate ratio estimate is represented by 𝑌̂𝑅𝑠 and is given as: 𝑌̂𝑅𝑠 =∑𝐿ℎ=1 ℎ 𝑋ℎ =∑𝐿ℎ=1 ℎ 𝑋ℎ , with the
𝑥ℎ 𝑥̅ ℎ
variance given in the following theorem.
Theorem 5.4: If an independent simple random sample is drawn in each stratum and sample sizes
are large in all strata, then the variance of 𝑌̂𝑅𝑠 is
𝑵𝟐 (𝟏−𝒇𝒉 )
𝑉(𝑌̂𝑅𝑠 ) = ∑𝐿ℎ=1 𝒉 2
(𝑠𝑦ℎ + 𝑅ℎ2 𝑠𝑥ℎ
2
− 2𝑅ℎ  𝑆𝑦ℎ 𝑆𝑥ℎ ), where Rh and h are the true ratio and
ℎ
𝒏𝒉
correlation in stratum h respectively.
The combined Ratio Estimate: In stratified sample, estimate of the population total Y is
𝑌̂𝑠𝑡 =N𝑦̅𝑠𝑡 =N∑ 𝑁ℎ 𝑦̅ℎ . For population total X, its estimate is
𝑋̂𝑠𝑡 =N𝑥̅𝑠𝑡 = N∑ 𝑁ℎ 𝑥̅ℎ .
𝑌̂ 𝑦̅
The combined ratio estimate 𝑌̂𝑅𝑐 , is given as 𝑌̂𝑅𝑐 =( 𝑠𝑡⁄̂ ) 𝑋=( 𝑠𝑡⁄𝑥̅ ) 𝑋,where 𝑦̅𝑠𝑡 and 𝑥̅ 𝑠𝑡 are the
𝑋𝑠𝑡 𝑠𝑡
estimated population means from a stratified sample, and X is known.
Theorem 5.5: If the total sample size n is large, the variance of 𝑌̅̂ is given by: 𝑅𝑐
W𝟐𝒉 (1−fh )
(𝑉(𝑌̅̂𝑅𝑐 ) = ∑𝐿ℎ=1 2
(𝑠𝑦ℎ + 𝑅ℎ2 𝑠𝑥ℎ
2
− 2𝑅ℎ ℎ 𝑆𝑦ℎ 𝑠𝑥ℎ ).
𝒏𝒉
𝟐
𝑵 (𝟏−𝒇 )
Corollary: 𝑉(𝑌̂𝑅𝑐 ) = ∑𝐿ℎ=1 𝒉 𝒏 𝒉 (𝑠𝑦ℎ
2
+ 𝑅ℎ2 𝑠𝑥ℎ
2
− 2𝑅ℎ ℎ 𝑆𝑦ℎ 𝑠𝑥ℎ ).
𝒉
𝑌 ∑𝑌 ̂ ̂
Corollary: Separate ratio estimate, 𝑅̂𝑠 , is given by 𝑅̂𝑠 = 𝑋𝑅𝑆 = 𝑋𝑅ℎ and its variance is
𝐿
V(𝑌̂𝑅𝑠 ) ∑ 𝑉(𝑌̂𝑅ℎ ) 1 𝑁ℎ2 (1 − 𝑓ℎ ) 2
𝑉(𝑅̂𝑠 ) = = = ∑ (𝑠𝑦ℎ + 𝑅ℎ2 𝑠𝑥ℎ
2
− 2𝑅ℎ ℎ 𝑆𝑦ℎ 𝑠𝑥ℎ )
𝑋2 𝑋2 𝑋2 𝑛ℎ
ℎ=1
𝐿
1 Wℎ2 (1 − 𝑓ℎ )
𝑉(𝑅̂𝑠 ) = ∑ 2
(𝑠𝑦ℎ + 𝑅ℎ2 𝑠𝑥ℎ
2
− 2𝑅ℎ ℎ 𝑆𝑦ℎ 𝑠𝑥ℎ )
𝑋̅ 2 𝑛ℎ
ℎ=1
𝑦̅ 𝑦̅ 𝑋 𝑌 ̅ ̅̂
Corollary: The combined ratio estimate for R is 𝑅̂𝑐 = 𝑥̅ 𝑠𝑡 = 𝑥̅ 𝑠𝑡 (𝑋̅) = 𝑋𝑅𝑐
̅
and its variance is
𝑠𝑡 𝑠𝑡

𝑉(𝑌̅̂𝑅𝑐 ) 1
𝐿
W𝒉𝟐 (1 − 𝑓ℎ ) 2
𝑉(𝑅̂𝑐 ) = = ∑ (𝑠𝑦ℎ + 𝑅ℎ2 𝑠𝑥ℎ
2
− 2𝑅ℎ ℎ 𝑆𝑦ℎ 𝑠𝑥ℎ )
𝑋̅ 2 𝑋̅ 2 𝑛ℎ
ℎ=1
Note: The separate ratio estimator is more efficient if Rh, population stratum ratio, varies
considerably and the sample size is large enough in each stratum. It is unlikely to improve the
efficiency of estimators if the same auxiliary variable is used for stratification and then for the ratio
method of estimation. If the population parameters are unknown in above expressions, substitute
the appropriate sample estimators for the parameters, i.e. , 𝑅̂ℎ , 𝑅̂ , 𝑠𝑦ℎ
2 2
, 𝑠𝑥ℎ , 𝑆𝑥𝑦ℎ , ̂ ℎ .

Sampling Theory | Chapter FIVE | Ratio Estimators 6