Study Notes on Sampling Methods & Surveys

Questions often arise regarding large populations, but gathering information from the entire population can be impractical or impossible.
- Instead, information is gathered from a smaller subset (sample) of the population.
Understanding the relationship between the population, sample, and inference is crucial.

Goal of Sampling: Select a sample that resembles the population as closely as possible, but in a smaller size.

Stratified Random Sample:
- Strata are groups that are similar within each group but not sampled entirely from all strata.
- Example: When tasting a Boston Cream pie, randomly select bites from the cake, custard, and frosting, ensuring representation of each layer.
Cluster Sample:
- Clusters differ within each group but are similar across groups; entire groups are sampled instead of individual members.
- Example: Randomly select a vertical slice of the pie to taste all layers at once.

Scenario: Estimating the average annual income of U.S. families from all 120 million households.
- Simple Random Sample: Randomly select households from the full list.
- Systematic Sample: Choose households at regular intervals (e.g., every 60,000th household).
- Stratified Random Sample: Divide households by census regions and take a random sample from each.
- Cluster Sample: Select entire census regions and include all households within the selected regions.

A visual representation highlighting different census regions and their subdivisions.

Convenience Sample: Obtained from individuals who are easiest to access, leading to possible misrepresentation of the broader population.
- Example: Asking friends about their favorite hockey teams, which may not reflect the views of the total fanbase.
Voluntary Sample: Participants respond to a general invitation; these samples can lead to underrepresentation in survey results, introducing bias.

Sampling Bias: A systematic failure of sampling to represent the population accurately.
Non-sampling Bias:
- Nonresponse Bias: Occurs when a significant number of selected participants do not respond to the survey.
- Response Bias: Any aspect of the survey that influences participants' responses.

Scenario: A market research firm asked, "What is your favorite soap?" which led to ambiguous results due to an unclear term.
- To improve, the question should clarify context (e.g., specify dish soap).

Scenario: Respondents are asked their opinion on sensitive topics like the death penalty using a 1-5 scale.
- Respondents may avoid extremes, often choosing middle options. To eliminate this bias, force respondents to take a side by asking for a binary response (support/oppose).

Scenario: Age categories in a poll (e.g., 30-40, 40-50) overlap, causing potential confusion.
- Revision needed to create distinct ranges, such as 30-39 and 40-49.

The methodologies and designs of surveys carry ethical responsibilities, particularly regarding bias and transparency of intentions in data collection.
Ensuring accurate representation and minimizing bias is fundamental to ethical standards in research, which contributes to reliable data interpretation and conclusions.

Understanding the various sampling methods, their biases, and the implications of survey design is essential for accurate statistical analysis and interpretation. Proper methodologies lead to trustworthy data that inform policy, research, and decision-making processes.