Sampling_Inference

Overview

Sampling, Surveys, and Statistical Inference is a medium-to-hard conceptual topic in the Problem-Solving and Data Analysis domain on the digital SAT. Questions test the ability to evaluate whether a sample is representative of a population, identify sources of bias in data collection, interpret margin of error and confidence intervals, and distinguish between what can and cannot be concluded from observational studies versus experiments. These are calculator-active questions, but they primarily require logical reasoning rather than computation.

Key Points

1. Random Sampling Types

Type	Description	SAT relevance
Simple random sample	Every member of the population has an equal chance of selection	Gold standard for generalization
Stratified random	Population divided into subgroups (strata); random sample taken from each	Ensures representation of subgroups
Cluster	Population divided into clusters; entire clusters selected randomly	Efficient for geographically dispersed populations
Systematic	Every nth member selected from a list	Practical; can generalize if list has no pattern

Key principle: A random sample is required to generalize findings to the broader population. A non-random (biased) sample cannot support general conclusions.

2. Sources of Bias

Bias Type	Description	Example
Voluntary response	People choose to respond; tend to have extreme opinions	Online polls, “call in” surveys
Convenience	Sample taken from whoever is easily accessible	Surveying only students in one classroom
Leading question	Question wording steers respondents toward a particular answer	”Don’t you agree that…?”
Undercoverage	Some population segments are systematically excluded	Phone surveys missing those without phones
Nonresponse	Selected individuals do not respond; pattern differs from respondents	Low survey return rates

All biased samples produce estimates that do not accurately represent the population.

3. Margin of Error and Confidence Intervals

A confidence interval gives a range within which the true population parameter likely falls.

$\text{Confidence Interval}=\hat{p}\pm \text{MOE}$

Where MOE = margin of error and p̂ = sample estimate.

Example: “42% ± 3%” → the interval is [39%, 45%].

The SAT always uses a 95% confidence level. This means: if the same survey were repeated many times, approximately 95% of the resulting intervals would contain the true parameter.

Effect of sample size on MOE:

$\text{MOE}\propto \frac{1}{\sqrt{n}}$

Larger sample → smaller MOE → more precise estimate. Smaller variability + larger sample → most precise estimate.

4. Observational Studies vs. Experiments

Feature	Observational Study	Experiment
Researcher role	Observes without intervening	Assigns subjects to conditions
Random assignment	No	Yes (in a well-designed experiment)
Can establish causation	No — association only	Yes — if random assignment used
Can generalize to population	Only if sample was random	Only if sample was random

Confounding (lurking) variable: A third variable correlated with both the explanatory and response variables, potentially explaining the observed association without a causal relationship.

Control group: Receives no treatment (or a placebo). Comparison to the treatment group isolates the effect of the intervention.

5. Valid and Invalid Conclusions — SAT Decision Tree

Was the sample randomly selected?
├── YES → Can generalize to the population
└── NO  → Cannot generalize; results apply only to the sample

Was there random assignment to groups?
├── YES (experiment) → Can claim causation
└── NO (observational) → Association only; cannot claim causation

6. Common SAT Inference Scenarios

Scenario	Valid conclusion	Invalid conclusion
Random survey of 500 city residents	Estimate city-wide preferences	Estimate national preferences
Study with random assignment, treatment shows improvement	Treatment causes improvement	Nothing if sample was non-random
Convenience sample shows trend	Trend exists in the sample	Trend applies to the population
Two variables are correlated in an observational study	Association exists	Causation

Pitfalls and Common Mistakes

Mistake 1: Concluding causation from an observational study. A study shows that people who eat breakfast have higher GPAs — students conclude breakfast causes higher GPA. Fix: Observational studies cannot establish causation. Look for confounding variables (e.g., students who eat breakfast may also have more structured home environments).

Mistake 2: Generalizing from a biased or non-random sample. A survey of website users is used to draw conclusions about “all adults.” Fix: Only random samples of the target population support generalization to that population.

Mistake 3: Misinterpreting the confidence interval. “95% confidence interval” means there is a 95% chance the true value is in this interval for a specific sample. Fix: The correct interpretation is: “We are 95% confident this procedure produces intervals that contain the true parameter.” The true parameter is fixed; the interval varies across samples.

Mistake 4: Thinking a larger margin of error is better. A larger MOE means less precision. Fix: Smaller MOE = better. To reduce MOE, increase sample size.

Mistake 5: Confusing the scope of a conclusion. A study on students at one school cannot be used to make inferences about all students nationally. Fix: Conclusions are only valid for the population from which the random sample was drawn, not a broader group.

Related Entries

• Probability
• Statistical_Measures
• Data_Distribution_Graphs
• Scatterplots_Regression
• Ratios_Proportions

Quick Reference Card

Concept	Rule
Random sample	Required to generalize to the population
Random assignment	Required to claim causation
Observational study	Association only — never causation
Experiment	Can establish causation if random assignment used
MOE interpretation	True value is within estimate ± MOE (95% confidence)
Larger n	Smaller MOE — more precise
Voluntary response bias	Overrepresents strong opinions
Convenience bias	Not representative; easy-to-access sample
Confounding variable	Third variable explaining apparent association
SAT confidence level	Always 95%