Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Scribd
Upload
19 views

Sampling and Sample Designsnew2016

Here are the key steps to draw a systematic sample: 1. Determine the sampling interval (K) by dividing the population size (N) by the desired sample size (n). In this case, K = 35,000/700 = 50. 2. Select the first record at random from the first K records listed in the sampling frame. 3. Then select every Kth record thereafter until the desired sample size of 700 is reached. So in summary, with a population of 35,000 and a desired sample of 700, the sampling interval would be every 50th record starting from a randomly selected first record.

Uploaded by

Ebaa Abdullatif
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
19 views

Sampling and Sample Designsnew2016

Here are the key steps to draw a systematic sample: 1. Determine the sampling interval (K) by dividing the population size (N) by the desired sample size (n). In this case, K = 35,000/700 = 50. 2. Select the first record at random from the first K records listed in the sampling frame. 3. Then select every Kth record thereafter until the desired sample size of 700 is reached. So in summary, with a population of 35,000 and a desired sample of 700, the sampling interval would be every 50th record starting from a randomly selected first record.

Uploaded by

Ebaa Abdullatif
Copyright
© © All Rights Reserved
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 36

Sampling and sample designs

Lecture 6
Survey research
• What do we need in order to conduct a
study?
-study problem
- Question
- Population- sample
- Methods
- Data analysis
- conclusions
STUDY POPULATION
Defining, Selecting and Sampling
Thinking about the
study population

• WHO?
• WHERE?
• WHEN?
• SAMPLE OR ENTIRE POPULATION?
Factors To Consider
APPROPRIATENESS
suitability to attain study objectives
»Volunteers
»Hospital patients
»Diagnosed patients
»People living at home
»Occupation groups
»Behavior-specific groups
»Autopsy reports
»Medical records
Factors To Consider

ACCESSIBILITY & PRACTICABILITY


will information be obtainable
from the whole study population?
» Cooperation
» Legal Access
» Mobility
» Sufficient numbers of individuals
Case Selection

Cases must be well-defined based on


detailed protocol of inclusion criteria
(avoids misclassification)

Caution: Too-stringent classification


may result in exclusion of cases who
did not meet criteria. But, too-loose
classification may result in inclusion
of non-cases
Selection of controls
1. Persons without the outcome whose
exposure to the (risk) factor of interest is
representative of the exposure experience
of the population that is the source of
cases & controls
 2. Persons with similar prevalence of
correlates of exposure to the risk factor
as the general population
Selection of controls
Population-based controls
 random sample from the community/population,
friends, colleagues, neighbors, relatives of cases

Hospital/clinic-based
 patients in same hospitals as cases
illness not related to outcome being studied
illness not related to exposure being studied
Do I always need a control group?
Only if testing an hypothesis
 • estimate the risk of infant
mortality
• Effect of maternal influenza
 during pregnancy on the risk of
infant mortality
• Dropping babies on their heads
 is not good for their health
Main question to be answered:

Can the results be generalized?


Depends on:

• Representativeness
• Coverage
• Size
GENERALIZABILITY PROBLEMS

1. Refusal to participate
2. Exclusion/inclusion not per protocol
3. Loss to follow-up

these can result in dissimilarity


between eligible study population and
actual study population (participants)
Target
Parent
Eligibles
Contacted
Sample
Target Population:
stroke cases
Parent Population:
stroke cases treated @ Ramallah Hosp
Eligible Individuals:
stroke cases, hospitalized@Ramallah, alive,
60-75 yrs, male or female, etc.,
who do not meet exclusion criteria
Contacted:
unchanged telephone/address
Study Population:
eligible cases who remain alive,
contactable, and agree to participate
Participation Rate =
number of participants
number of eligibiles

Response Rate =
number of participants
number contacted
SAMPLING
Why sample?

• It is not always possible to collect


information from everyone
• Besides, information from a well-
selected sample may be as informative
as from the entire population

$
Sampling Design
Probability Sample Non-probability Sample
• Based on laws of chance • Based on human judgment
• Allows for error estimation • Cannot empirically
and likelihood of reflecting
estimate individual’s
parent population –
external validity chance of being sampled

Example: high-school survey on vandalism


Enter student ID#s into computer and Ask principal/teachers about which
generate a sample of every 10th student students are/are not vandals, who will
cooperate…
Sampling frame
A list—real or theoretical—of the population
from which sample is drawn
Underlying assumption of any study: sample
is representative of the population defined by
the sampling frame
This allows results obtained from sample data
to be generalized to the larger population
Types of Samples
Probability
° Simple Random
° Systematic Random
° Stratified
° Cluster Non-Probability
• Purposive
• Quota
• Convenience/Chunk
RANDOM
Each unit has equal probability to be selected
Steps:
1. Construct sample frame
RANDOM
Each unit has equal probability to be selected
Steps:

 
1. Construct sample frame
2. Decide upon sample size (sampling
fraction)              

     
RANDOM
Each unit has equal probability to be selected
Steps:
1. Construct sample frame
2. Decide upon sample size (sampling
fraction)
3. Select units at random
CLUSTER
Units are grouped & groups are selected
Steps:
1. Identify groups or arrange units into groups
2. Decide upon sample size
3. Select units randomly or
systematically
Multi-stage Sampling

Primary Secondary Sampling


units units units
STRATIFIED
Units are divided into strata
& sampled within strata
Steps:
1. Divide sample frame into relevant strata
2. Decide upon required sample size within
each stratum
3. Select units randomly or systematically
STRATIFIED
Why stratify?
• Enhance coverage and precision
How to draw a stratified sample
• The problem: in a study of revitalization in
an urban neighborhood. We plan to
examine the attitudes of new residents
toward their community. We anticipated
that the attitudes of homeowners may
differ from those of renters. So to ensure
the proper presentation of both groups, we
will use a proportional startified random
sample with two starta: new homeowners
and new renters.
How to draw a stratified sample
• 1. the population consists of N= N1 + N2, with
N1 denoting new homeowners and N2 new
renters. N1 = 200 and N2= 300. N=500. we
decide to select a proportional sampling fraction
of 1/10 from each stratum. Thus N1= 20
homeowners and N2= 30 renters will be
included in the sample.
• 2. Use simple random sampling separatly for
each list.
Sex-Age
Pop # Pop %
group
F, <65 17,881 49%

F, 65+ 1,825 5%

M, <65 15,326 42%

M, 65+ 1,460 4%

Total 36,492 100%


Sex-Age
Pop # Pop % Sample
group
F, <65 17,881 49%

F, 65+ 1,825 5%

M, <65 15,326 42%

M, 65+ 1,460 4%

Total 36,492 100% 2500


Sex-Age
# % Sample
group
F, <65 17,881 49%  1225

F, 65+ 1,825 5%  125

M, <65 15,326 42%  1050

M, 65+ 1,460 4%  100

Total 36,492 100% 2500


Proportional allocation
SYSTEMATIC
Every kth unit selected
Steps:
1. Construct sample frame
2. Decide upon sample size
3. Decide upon sampling interval (k)
SYSTEMATIC
Strictly speaking, systematic sampling
does not give every eligible unit an equal
chance of being selected, however in
practice, systematic random sampling is
accepted as random sampling (simple
random vs. random)
How to draw a systematic sample
• A social scientist is interested in investigating the
relationship between parent’s occupations and
the averages of students on a large urban
campus.(N=35,000). n=700 records will be
selected.
• 1. determine the samping interval, K. As N=
35,000 and n=700, k is 35,000/700= 50.
• 2. select the first record at random from the first
K=50 records listed and then select every 50
record thereafter until the sample size of 700 is
reached.

You might also like