3.4 - Small Population Example

Example 3-4: Wheat Production Section

( Reference Section 6.4 of the text)

unit (Farm) i
1
2
3
Selection Prob, \(p_i\)
0.3
0.2
0.5
Wheat produced
11
6
25

N = 3 farms, n = 2 sample with replacement.

s
p(s)
ys
1, 1
0.3(0.3) = 0.09
(11, 11)
2, 2
0.2(0.2) = 0.04
(6, 6)
3, 3
0.5(0.5) = 0.25
(25, 25)
1, 2
0.3(0.2) = 0.06
(11, 6)
2, 1
0.2(0.3) = 0.06
(6, 11)
1, 3
0.3(0.5) = 0.15
(11, 25)
3, 1
0.5(0.3) = 0.15
(25, 11)
2, 3
0.2(0.5) = 0.10
(6, 25)
3, 2
0.5(0.2) = 0.10
(25, 6)

Question: Compute the Hansen-Hurwitz estimator.

Answer

When (1,1) is sampled, the Hansen-Hurwitz estimator is:

\(\hat{\tau}_p=\dfrac{1}{2}\left(\dfrac{y_1}{p_1}+\dfrac{y_1}{p_1}\right)=\dfrac{1}{2}\left(\dfrac{11}{0.3}+\dfrac{11}{0.3}\right)=36.67\)

When (1,2) is sampled, the Hansen-Hurwitz estimator is:

\(\hat{\tau}_p=\dfrac{1}{2}\left(\dfrac{y_1}{p_1}+\dfrac{y_2}{p_2}\right)=\dfrac{1}{2}\left(\dfrac{11}{0.3}+\dfrac{6}{0.2}\right)=33.33\)

Similarly, we can fill out the table and get the Hansen-Hurwitz estimators as shown below:

s
p(s)
ys
\(\hat{\tau}_p\)
1, 1
0.3(0.3) = 0.09
(11, 11)
36.67
2, 2
0.2(0.2) = 0.04
(6, 6)
30.00
3, 3
0.5(0.5) = 0.25
(25, 25)
50.00
1, 2
0.3(0.2) = 0.06
(11, 6)
33.33
2, 1
0.2(0.3) = 0.06
(6, 11)
33.33
1, 3
0.3(0.5) = 0.15
(11, 25)
43.33
3, 1
0.5(0.3) = 0.15
(25, 11)
43.33
2, 3
0.2(0.5) = 0.10
(6, 25)
40.00
3, 2
0.5(0.2) = 0.10
(25, 6)
40.00

Question: Compute the Horvitz-Thompson estimator.

Answer

\(\pi_1=0.09+0.06+0.06+0.15+0.15=0.51\)
\(\pi_2=0.04+0.06+0.06+0.10+0.10=0.36\)
\(\pi_3=0.25+0.15+0.15+0.10+0.10=0.75\)

When (1,1) is sampled, the Horvitz-Thompson estimator is:

\(\hat{\tau}_\pi=\left(\dfrac{11}{0.51}\right)=21.57\)

When (1,2) is sampled, the Horvitz-Thompson estimator is:

\(\hat{\tau}_\pi=\left(\dfrac{11}{0.51}+\dfrac{6}{0.36}\right)=38.24\)

Similarly, we can fill out the table and get the Horvitz-Thompson estimators as shown below:

s p(s) ys
\(\hat{\tau}_p\)
\(\hat{\tau}_\pi\)
1, 1 0.3(0.3) = 0.09 (11, 11) 36.67 21.57
2, 2 0.2(0.2) = 0.04 (6, 6) 30.00 16.67
3, 3 0.5(0.5) = 0.25 (25, 25) 50.00 33.33
1, 2 0.3(0.2) = 0.06 (11, 6) 33.33 38.24
2, 1 0.2(0.3) = 0.06 (6, 11) 33.33 38.24
1, 3 0.3(0.5) = 0.15 (11, 25) 43.33 54.90
3, 1 0.5(0.3) = 0.15 (25, 11) 43.33 54.90
2, 3 0.2(0.5) = 0.10 (6, 25) 40.00 50.00
3, 2 0.5(0.2) = 0.10 (25, 6) 40.00 50.00
Mean     42 42
Variance     34.67 146.46

\(\text{Mean of }\hat{\tau}_p = E(\hat{\tau}_p)=\sum{p(s) \cdot \hat{\tau}_p (s)}\)
\(= 0.09\times36.67+0.04\times30.00+0.25\times50.00+0.06\times33.33+0.06\times33.33\)
\(+0.15\times43.33+0.15\times43.33+0.10\times40.00+0.10\times40.00\)
\(= 42\)

From the table above we can see that both \(\hat{\tau}_p\) and \(\hat{\tau}_\pi\)are unbiased. This example is a small population example to illustrate conceptually the properties of these estimators. We can compute the variance for \(\hat{\tau}_p\)and the variance for \(\hat{\tau}_\pi\) directly from the definition of variance.

Since mean of \(\hat{\tau}_p\) is 42 and E(g) = \(\sum p(s)*g(s)\), we can compute the variance of \(\hat{\tau}_p\) as:

\(Var(\hat{\tau}_p) = E[\hat{\tau}_p-\text{mean of }\hat{\tau}_p]^2\)
\(= 0.09\times(36.67-42)^2+0.04\times(30.00-42)^2+0.25\times(50.00-42)^2+0.06\times(33.33-42)^2\)
\( +0.06\times(33.33-42)^2+0.15\times(43.33-42)^2+0.15\times(43.33-42)^2+0.10\times(40.00-42)^2\)  \( +0.10\times(40.00-42)^2\)
\(=34.67\)

Similarly, we can compute that the variance of \(\hat{\tau}_\pi\)is 146.46.

Now, how do we compute the MSE of \(\hat{\tau}_p\)?

By definition, \( MSE (\hat{\tau}_p) = E(\hat{\tau}_p-\tau)^2\), in this case, since \(\hat{\tau}_p\)is unbiased and \(E(\hat{\tau}_p) =\tau\) ,\( MSE (\hat{\tau}_p) \)is the same as \(Var(\hat{\tau}_p)\).

Remark 1. The above demonstration is just a teaching tool. In reality, we will not know the population and will not come across small population problems like this other than in exams and assignments. What we know is:

Unit
1
2
3
Selection probability
0.3
0.2
0.5

And, we draw a sample. If the sample we draw is (1,2) then \(\hat{\tau}_p\) = 33.33 and\(\hat{\tau}_\pi=38.24\) .

We will not be able to find the real population total nor the real variance of the estimator. However, we will be able to estimate them.

Remark 2. Now, should we use \(\hat{\tau}_p\)or should we use \(\hat{\tau}_\pi\)?

There are no clear answers. Both estimators ar acceptable when \(y_i\) and \(p_i\) are proportional.