8.10 Sampling

Question 1

Sampling?

Answer 1

The process of selecting a subset of observations for the purpose of drawing conclusions about the larger set of observations.

Question 2

Population?

Answer 2

An aggregation of all cases to which research findings are generalized

Question 3

Sample?

Answer 3

A portion of a population selected for study

Question 4

Key Principle of sampling?

Answer 4

Representativeness

• The sample has approximately the same characteristics of the population relevant to the research in question.
• Selection of the few who can be taken to represent the many.

Question 5

Reasons for Sampling?

Answer 5

Practical considerations
Accuracy
Efficiency

Question 6

Target Population?

Answer 6

*Must be clearly identified
Unit of observation
Locale
Time
Characteristics

Question 7

Target Population Examples?

Answer 7

All 12th grade students in public schools in school year 1999-2000 in Taiwan.
All women aged 65 and older living in long-term care facilities in Taipei.

Question 8

Sampling Frame?

Answer 8

Actual list or definition of all cases from which the sample is selected

Question 9

Sampling Designs?

Answer 9

-Probability
Cases are selected using the process of random selection.
Chances of selecting a case are known.
-Nonprobability
Cases are selected using other means than random selection.
Chances of selecting a case are not known.

Question 10

Probability Sampling Designs?

Answer 10

• Simple Random Sampling
• Systematic Sampling
• Stratified Random Sampling
• Cluster Sampling

Question 11

Advantages of Probability Sampling Designs?

Answer 11

• Avoids bias in selection of cases.
• Because all cases have a known probability of being selected, the chances are excellent that the sample selected will closely resemble the population.
• Permits the application of sampling probability theory for estimating the sample accuracy.

Question 12

Simple Random Sampling?

Answer 12

• Basic probability sampling design
• Every possible combination of cases has an equal probability of being selected
• Requires a complete list of the population and random selection of cases

Question 13

Simple Random Sampling Steps?

Answer 13

1. Define the target population.
2. List or define all cases in the sampling frame.
3. Assign identification number to each case in the sampling frame.
4. Use a random criterion to select cases (by ID) for the sample (e.g., random number table).

Question 14

Systematic Sampling?

Answer 14

Selection of every Kth case from the sampling frame with a random start from the first K cases on the list.
K = sampling interval
Caution: possibility of bias if periodic or cyclical patterns are present in the frame.

Question 15

Systematic Random Sampling Steps?

Answer 15

1. Define the target population.
2. List or define all cases in the sampling frame and assign IDs.
3. Divide the size of the sampling frame by the sample size to identify the sampling interval.
4. Identify the starting point by randomly selecting one case from the set of cases in the first interval.
5. With that case as starting point, select every Kth case on the list.

Question 16

Stratified Random Sampling?

Answer 16

-Population subdivided into strata
-Simple random samples drawn from each strata
-Can produce a better (more precise) sample if the stratifying variable(s) is related to the dependent variable.
Because if the stratifying variable is strongly related to the estimated variable, much of the variation in the estimated variable can be explained by the differences between the strata. Then we reduce a large proportion of variation.

Question 17

Stratified Random Sampling ex & note?

Answer 17

-For example, with respect to alcohol drinking, there are large variability between women and men. If we conducted stratified random sampling by gender, we can eliminate this “between gender” source of heterogeneity, leaving homogeneity of variance in each gender. Then we only require a relatively small sample.
-Consider “cost of stratifying” vs. “cost of sampling more” cases
If you want to stratify by gender, you need to know prior to drawing the sample whether each person is a male or a female

Question 18

Stratified Random Sampling Steps?

Answer 18

1. Define the target population.
2. List or define all the cases in the sampling frame and assign Ids.
3. Identify the stratifying variables and subdivide/sort the sampling frame based on those variables.
4. Use simple or systematic random sampling to draw cases from each stratum.
5. Combine the cases from the strata to form the study sample.

Question 19

Stratified Random Sampling Example (Proportionate)?

Answer 19

Race distribution in sample of 10,000
	African-American	12%      1,200   
    American Indian	  1%         100
	Asian or Pac. Isl.	  2%         200   
	White	           	80%       8,000
	Other		            5%          500

Question 20

Stratified Random Sampling Example (Disproportionate)?

Answer 20

Race distribution in sample of 10,000
	African-American	 20%       2,000   
    American Indian	 20%       2,000
	Asian or Pac. Isl.	 20%       2,000   
	White	           	 20%       2,000
	Other		            20%       2,000
Disproportionate random sampling requires statistical adjustment (weighting) to make generalization to the population.

Question 21

Cluster Sampling?

Answer 21

• Population is broken into clusters and clusters are randomly selected.
• In single stage cluster sampling, all cases in the selected clusters are included.
• In multistage cluster sampling, sampling occurs at two or more stages.

Question 22

Cluster Sampling Steps?

Answer 22

1. Define the target population and sampling frame of clusters.
2. Use simple or stratified random sampling to select whole clusters (single-stage cluster sampling stops here; use all cases in selected clusters).
3. Use simple or stratified random sampling to select individual cases (or another stages of clusters) from the clusters selected in step 2 (multistage sampling)

Question 23

Purpose of Using Cluster sampling?

Answer 23

-While stratified random sampling is used either to increase sample precision or to provide a sufficient number of cases in small strata, the principal reason for cluster sampling is to reduce the costs of data collection
Interviewer travel and the listing of population elements

Question 24

Precision of Cluster Sampling?

Answer 24

Low cost but bad efficiency in precision
-In two-stage cluster sampling, there is variability both between and within the clusters.
-Sample more clusters or. more people?
Consider the heterogeneity between clusters and the heterogeneity between people
It is dangerous to sample only a few clusters, because it cannot account for the differences between clusters
-The more stages, the larger the total sampling error tends to be.

Question 25

Probability Sampling Designs Example?

Answer 25

Target population: alcohol beverage (beer/wine) retailers in Taipei
Sampling frame: 80,000 beverage retailers identified on name and address lists maintained by Taipei government
Sample size: 800 retailers

Question 26

Example con’t

Answer 26

-Simple random sampling
Randomly select 800 retailers using a random number table (ignore repeats).
-Systematic random sampling
Calculate sampling interval (80,000/800=100); randomly select starting point from first 100 cases; select every 100th case from there.
-Stratified random sampling
Identify relevant stratifying variables: Type outlet (restaurant , pub, convenient store)
Subdivide list into strata; randomly select from each strata; combine cases to form single sample
-Cluster sampling
Identify primary sampling units and secondary sampling units (retail outlets).
Randomly select clusters; use all cases in each selected cluster (single stage cluster sampling) or randomly select cases from each cluster (multistage cluster sampling); combine cases from clusters to form a single sample.

Question 27

Nonprobability Sampling Designs?

Answer 27

• Convenience (accidental) - cases selected from those conveniently available.
• Purposive - cases selected that are judged to be representative of the population.
• Quota - cases judged to be representative selected within strata up to a fixed quota.

Question 28

Weaknesses of Nonprobability Sampling?

Answer 28

No control for investigator bias in selecting cases (selection bias)
Pattern of variability cannot be predicted from sampling probability theory, therefore impossible to predict the chances that the sample and population distributions are similar.
Not equal to “purposeful sampling” in qualitative studies

Question 29

Sample Accuracy?

Answer 29

Sampling error: The amount a given sample statistic deviates from the population parameter.
Example:
Population mean = 3.00
If we draw a sample, which has mean = 2.50
then sampling error = .50

Question 30

Implications for Sample Accuracy of Sampling Error?

Answer 30

As samples become larger and more homogeneous, sampling error decreases.
As sampling error decreases, the sample is more likely to be representative of the target population.
The more representative a sample, the more generalizable to the population the results.