Cluster Sampling and Types of Bias in Statistics

Study Guide - Smart Notes

Tailored notes based on your materials, expanded with key definitions, examples, and context.

Cluster Sampling

Definition and Process

Cluster sampling is a probability sampling technique commonly used in statistics when it is difficult or impractical to obtain a complete list of the population. It involves dividing the population into groups, called clusters, and then randomly selecting some clusters to include all individuals from those clusters in the sample.

Step 1: Divide the population into two or more non-overlapping groups, called clusters.
Step 2: Randomly choose some of the clusters.
Step 3: Include all individuals in the chosen clusters in the sample.

Example

A researcher in a large city wants to determine the prevalence of suspensions among fifth-graders. She does not have a list of all fifth-graders, but she does have a list of all 60 elementary schools in the city. She treats each school as a cluster, randomly selects 10 schools, and requests the suspension history of all fifth-graders in those schools. This is an example of cluster sampling.

Practical Application: Selecting Clusters

Suppose you are the researcher and need to select 10 schools out of 60 (numbered 01 to 60). Using a Table of Random Numbers, you randomly select the following schools: 03, 36, 55, 04, 47, 51, 22, 59, 37, 16. You would then collect data from all fifth-graders in these schools.

Advantages of Cluster Sampling

No need for a complete frame: In some cases, a list of clusters (e.g., schools, apartment buildings, city blocks) is sufficient.
Cost-effective: Reduces costs by focusing on selected clusters, saving travel and time expenses.

Disadvantages of Cluster Sampling

Risk of unrepresentative samples: If individuals within clusters are too similar and clusters differ as aggregate units, the sample may not represent the population well.
Implementation challenges: It can be difficult to identify and define clusters appropriately.

Bias in Sampling

Definition of Bias

A sample is said to have bias if its characteristics are not representative of the population. Bias can distort statistical conclusions and lead to invalid results.

Types of Bias

Sampling bias
Nonresponse bias
Response bias

Sampling Bias

Sampling bias occurs when the sampling technique favors one part of the population over another. This often results from undercoverage, where the sampling frame omits a segment of the population.

Example

If a survey uses a list of households with telephones, it excludes those without phones, potentially missing opinions that differ from those included.

Nonresponse Bias

Nonresponse bias arises when individuals selected for the sample do not respond, and their opinions or characteristics differ from those who do respond.

Mitigation strategies: Use callbacks (follow-up calls or visits) and incentives (coupons or cash rewards) to increase response rates.

Response Bias

Response bias occurs when there is a tendency for individuals to answer survey questions incorrectly or falsely. Several sources contribute to response bias:

Interviewer error: The interviewer may influence answers through their behavior or tone.
Misrepresented answers: Respondents may provide inaccurate or untruthful responses, often to present themselves favorably.
Wording of questions: Leading, double-barreled, or vague questions can confuse respondents or influence their answers.
Ordering of questions and words: The sequence of questions or the order of answer choices can prime respondents and affect their responses.

Examples of Response Bias

Leading question: "Are you in favor of the construction of a new shopping center, which will result in new jobs?"
Double-barreled question: "Do you agree that this detergent smells good and removes all stains?"
Vague question: "How much do you exercise?" (Better: "How many hours do you spend exercising each week?")

Summary Table: Types of Bias

Type of Bias	Definition	Example
Sampling Bias	Sample favors one part of the population due to undercoverage	Survey using only households with telephones
Nonresponse Bias	Individuals who do not respond differ from those who do	Low response rate in mailed surveys
Response Bias	Respondents answer incorrectly or falsely	Leading questions, interviewer influence

Key Terms and Concepts

Cluster sample: A sample where entire groups (clusters) are randomly selected and all members of those groups are included.
Bias: Systematic error that leads to samples not representing the population.
Undercoverage: When some groups in the population are not included in the sampling frame.

Formulas

While cluster sampling does not have a specific formula, the general probability of selecting a cluster can be expressed as:

For bias, there is no direct formula, but the concept is crucial in designing representative samples.

Additional info: Academic context and examples have been expanded for clarity and completeness.