01 / 20

Data Collection
Methods

Master the key techniques for gathering reliable data: surveys, experiments, observational studies, and how to identify bias in each approach.

ACT Math

What Is Data Collection?

Data collection is the systematic process of gathering information from a defined population or sample. The method you choose directly affects the validity and reliability of your conclusions.

Essential Vocabulary

Population
The entire group you want to study or draw conclusions about.
Sample
A subset of the population actually measured or surveyed.
Bias
A systematic error that makes results unrepresentative of the population.

Three Core Methods

Survey Option A Option B Option C Ask questions Experiment Control Treatment Compare Impose treatment Observational Record No intervention

Surveys and Sampling

A survey collects self-reported data through questions. The critical factor is how the sample is chosen. A well-designed survey uses random selection so that every member of the population has an equal chance of being included.

1
Simple Random Sample (SRS): Every individual has an equal probability of selection. Gold standard for unbiased results.
2
Stratified Sample: Divide the population into subgroups (strata), then randomly sample within each. Ensures representation.
3
Systematic Sample: Select every \( k^{\text{th}} \) individual from an ordered list (e.g., every 10th person). Efficient and often representative.
4
Convenience Sample: Choose whoever is easiest to reach. High risk of bias: results may not generalize.

Experiments and Causation

An experiment deliberately imposes a treatment on subjects to measure its effect. Only experiments can establish cause-and-effect relationships.

Control Group
Receives no treatment or a placebo. Serves as the baseline for comparison.
Treatment Group
Receives the intervention being tested. Differences from control reveal the effect.

Observational Studies

In an observational study, researchers record data without imposing any treatment. They observe subjects in their natural state or review existing records.

1
No intervention: The researcher does not manipulate any variable.
2
Can show association: Observational data can reveal correlations between variables.
3
Cannot prove causation: Confounding variables may explain the relationship.

Causation vs. Correlation

Experiment Cause X Effect Y X causes Y (treatment imposed) ✓ Observational Variable X Variable Y Confounder? Correlation only (no treatment imposed)

Sources of Bias

Selection Bias
The sample systematically excludes part of the population. Example: surveying only gym members about exercise habits.
Response Bias
Participants give inaccurate answers, often due to question wording or social pressure.
Nonresponse Bias
Those who don't respond differ meaningfully from those who do. Low response rates increase this risk.

How to Evaluate a Study

1
Identify the method: Is it a survey, experiment, or observational study?
2
Check the sample: Was it random? Does it represent the population?
3
Look for bias: Could selection, response, or nonresponse bias affect results?
4
Evaluate the conclusion: Does the method support the claim? Causation requires an experiment.
5
Scope of inference: Can results generalize to the whole population, or only to the sample?

Example 1: School Survey

1
Method: Survey (self-reported responses via questionnaire).
2
Sample: Simple random sample from the full enrollment list: representative.
3
Conclusion scope: Results can generalize to all students in the school.
4
Limitation: Cannot prove a later start time causes better outcomes: only measures preference.

Example 2: Drug Trial

Method
Randomized controlled experiment with a placebo control group.
Conclusion
If the treatment group improves significantly, we can infer the drug caused the improvement.

Example 3: Coffee Study

1
Method: Observational study (no treatment imposed).
2
Conclusion: There is an association between coffee and heart disease.
3
Not valid: "Coffee causes heart disease." Heavy coffee drinkers may also smoke, sleep less, or have higher stress: all confounders.

Common Mistakes

✕ Claiming causation from observation
Saying "X causes Y" when no treatment was imposed. Observational studies show correlation only.
✕ Ignoring sample method
Treating a convenience or voluntary sample as if it were random. Results cannot generalize to the population.
✕ Confusing sample and population
Applying findings from a specific group to everyone. Results only generalize to the population the sample was drawn from.

Comparison Summary

Survey Experiment Observational
Treatment imposed? No Yes No
Can show causation? No Yes No
Data source Self-report Measured outcomes Recorded behavior
Main risk Response bias Ethical limits Confounders
Generalize? If random sample If random sample If random sample

Question 1 of 5

A researcher randomly assigns 100 volunteers to either take a vitamin supplement or a placebo for 12 weeks, then measures their energy levels. What type of study is this?

Question 2 of 5

A study finds that students who eat breakfast daily have higher test scores than those who skip breakfast. The researchers did not assign eating habits. Which conclusion is valid?

Question 3 of 5

A magazine asks readers to mail in their opinion on a proposed tax increase. 72% of respondents oppose the tax. Which type of bias is most likely present?

Question 4 of 5

In a well-designed experiment, what is the primary purpose of random assignment to treatment groups?

Question 5 of 5

A school surveys students by selecting every \( 10^{\text{th}} \) name from an alphabetical roster. This sampling technique is called: