RELATIVE PRECISION OF STRATIFIED RANDOM SAMPLING WITH SIMPLE RANDOM SAMPLING
RELATIVE PRECISION
OF STRATIFIED RANDOM SAMPLING
WITH
SIMPLE RANDOM
SAMPLING
-Garima Singh
Here
we are mainly focusing on a comparative study of SRSWOR & Stratified Random
Sampling under various methods of allocations.
Let’s
start with a simple introduction.
What
is sampling?
For
the purpose of determining population characteristics, instead of enumerating
entire population, we take a finite subset of statistical individuals from the
population called a sample and this process through which the required number
of units are selected is called sampling.
What
is Simple Random Sampling (SRS)?
In
SRS, we draw a sample in such a way that each unit of the population has an
equal & independent chance of being selected in the sample. It is of two
types:
1.
Simple
Random Sampling With Replacement (SRSWR)
2.
Simple
Random Sampling Without Replacement (SRSWOR)
What
is Stratified Random Sampling?
In
Stratified Random Sampling, the whole population is divided into homogeneous
groups under certain criterion. These groups are termed as strata.
Then
the sample is drawn randomly from each stratum independently. The estimates are
then calculated from the data obtained from all the strata. The allocation of
the sample sizes for different strata is done in the following ways:
a)
Equal
Allocation
nh=n/L
where nh
is the sample size of hth stratum
L is
the total no. of stratums
n is
the sample size
b)
Proportional
Allocation
nh= (n/N) Nh
where Nh is the stratum size of hth
stratum
c)
Optimum
Allocation
nh= n (Nh Sh)/∑ (Nh Sh)
Well
now we have some information before proceeding further.
How
to decide which estimator is more efficient?
Suppose
we want to compare which estimator is more efficient, one which is obtained
using SRSWOR or one which is obtained using the Stratified Random Sampling to
estimate, let’s say population mean. One way to answer this is to compare the variance
of the estimates.
Lesser the variance, more is the
efficiency then more precise is the estimate.
Let’s
take an example to understand how we are using the above-mentioned idea to
decide which estimator gives the more precise estimate for population mean.
Example-
A
sample survey is to be undertaken to ascertain the mean annual income of farms
in a certain area. The farms are stratified according to their principal
products. A census conducted gave the following results:
Type of farm |
Number of farms |
Mean annual
income |
Standard
deviation |
Sheep |
161 |
10946 |
2236 |
Wheat |
195 |
6402 |
2614 |
Dairy |
274 |
2228 |
606 |
Others |
382 |
1458 |
230 |
For
a sample of 120 farms compute the sample sizes in each stratum under
proportional and optimum allocation. Compare the efficiency of these methods
with that of the SRSWOR by telling which scheme is more efficient and gives
more precise estimate of population mean i.e. the mean annual income.
Important
Formulae used:
1.
Variance
of estimate of population mean using SRSWOR:
where
2.
Variance
of estimate of population mean using Stratified Random Sampling under
proportional allocation:
3.
Variance
of estimate of population mean using Stratified Random Sampling under
proportional allocation:
Variance of the estimates using R Codes:
Hence,
we can see from the above R codes that,
Vsrswor>Vprop>Vopt
This
implies that the estimated mean of Annual Income obtained by using Stratified
Random Sampling under optimum allocation is most efficient compared to other
two schemes hence it gives the most precise estimate for estimating the mean
annual income of farms in a certain area.
So
now we understand with this example if we have some information about the population,
we can easily verify which estimator we should use to estimate the population
parameter.
Now you can take any example, any problem and see for yourself which sampling scheme you should take!
I hope this blog helped you to understand why stratified random sampling gives more precise estimator compared to SRSWOR.
Comments
Post a Comment