Sampling formulae

Simple Random Sampling

Estimate of Population Mean 

(4.1)

Estimate of variance of mean

(4.2)

Where

Bound on the error of estimation = (4.3)

Estimator of the population total  (SRS)

(4.4)

Estimated variance of 

(4.5)

Bound on the error of estimation

(4.6)

Sample size required to estimate with a bound on the error of estimation B:

(4.11)

where D=

1
Sample size required to estimate with a bound on the error of estimation B:

(4.13)

where D=

Estimation of Population Proportion based on SRS

yi = 1if the element possess a certain characteristic, 0 if not.

(4.14)

Estimated variance of :

(4.15)

Bound on the error of estimation

(4.16)

Sample size required to estimate p with a bound on the error of estimation B:

(4.18)

where q=1-p dan D =

Stratified Random Sampling

Estimator of the population mean

2
 (5.1)

Estimated variance of

(5.2)

Estimator of the population total 

(5.3)

Estimated variance of the population total  =

= (5.4)

ni = n x ai where ai is the fraction of observations allocated to stratum i. (5.5)

Approximate sample size required to estimate or with a bound B on the error of estimation.

(5.6)

where ai = ni/n
and D= B2 /4 for estimating mean 
D=B2 /4N2 for estimating total 

Sample allocation

1. Equal allocation

ai= 1/number of stratum

2. Proportional allocation

3
ai= Ni/N

3. Neyman allocation

(5.9)

4. Optimum allocation
Approximate allocation that minimizes cost for a fixed value of V ( ) or minimum
V( ) for a fixed cost:

(5.7)

Estimator of the population proportion p:

(5.13)

Estimated variance of

(5.14)

Approximate sample size required to estimate p with a bound B on the error of estimation:

(5.15)

Approximate allocation that minimizes cost for a fixed value of or minimizes

for a fixed cost:

(5.16)

Systematic Sampling

4
Estimator of the population mean

(7.1)

Estimated variance of

(7.2)

Estimator of the population total

(7.5)

Estimated variance of

(7.6)

Estimator of the population proportion p

(7.7)

Estimated variance of

(7.8)

Sample size required to estimate p with a bound B on the error of estimation:

(7.10)

where

Sample size required to estimate p with a bound B on the error of estimation:

5
 (7.11)

where q=1-p and

Cluster Sampling

Estimator of the population mean

(8.1)

Estimated variance of

(8.2)

where

(8.3)

Estimator of the population total

(8.4)

Estimated variance of

(8.5)

If M is not known, then

6
 (8.6)

Estimator of the population , which does not depend on M:

(8.7)

Estimated variance of N

(8.8)
where

(8.9)

Approximate sample size required to estimate , with a bound B on the error of estimation:

(8.12)

where is estimated by and D=

Approximate size required to estimate , using M , with a bound B on the error of estimation.

(8.13)

where is estimated by , and

Approximate sample size required to estimate , using , with a bound B on the error of
estimation:

(8.15)

where is estimated by and

Estimator of the population proportion p:

7
 (8.16)

Estimated variance of

(8.17)

where

(8.18)

Cluster sampling with probabilities proportional to size

Estimator of the population mean :

(8.19)

Estimated variance of

(8.20)

Estimator of the population total :

(8.21)

Estimated variance of

(8 .22)

Two-Stage Cluster Sampling

Unbiased estimator of the population mean µ

8
 (9.1)

Estimated variance of

(9.2)

(9.3)

i=1, 2,3,’’’’n (9.4)

Estimation of the population total 

( 9.5)

Estimated variance of

(9.6)

Ratio estimator of the population mean µ

(9.7)

Estimated variance of

(9.8)

where

9
 (9.9)

and

n=1,2,.....n (9.10)

Estimator of a population proportion p:

(9.11)

Estimated variance of p:

(9.12)

where

(9.13)

Sampling equal-size clusters

(9.14)

9.2 now becomes

(9.15)

where f1= n/N and

10
When N is large
(9.16)

(9.17)

where = variance among the cluster means

= variance among the elements within clusters

(9.19)

where c1 is associated with cost of sampling each cluster and c 2 is associated with cost of
sampling each element within a cluster, and c is the total cost.

The value of m that minimizes for fixed cost, or minimizes c for fixed variance is

(9.20)

Two-Stage Cluster Sampling with Probabilities Proportional to Size

Estimator of population mean µ

(9.22)

Estimated variance of

(9.23)

Estimator of the population total

(9.24)

Estimated variance of

11
 (9.25)

Control of Sub-sample size

f=

b= number of cases that we want to select, and N i is the cluster size for the i th cluster, f is the
sampling fraction and F is the inverse of f.

[N / F x b] x [b x N] = 1/F = f

First stage second stage overall fraction sampling fraction

Design effects
_ _
Var (yst) / Var (y srs) for stratified random sample, and its value is less than 1 if the design is more
efficient than the srs
_ _
Var (ycl) / Var (ysrs) for cluster sampling

If one has the estimate of the design effect, and the variance from srs, one can estimate the
variance from complex sample, simply by
_ _
Var (ycl) = deft2 [Var (ysrs)]
_
Deft2 = 1+  (n-1)
_
 = [deft2 -1]/ [n-1]

where  =intraclass homogeneity

_
n = average number of elements in the selected cluster

This implies that if the sample size if srs is reduced by half, the varians will be double, given
_
Var (y) = s2 / n
_ _
Thus, if s2 is constant, Var (y) =constant/100 will be that of Var (y) = constant/200.

12