AP Statistics common mistakes
AP Statistics · Understand the traps and how to avoid them.
Describing the shape, center, and spread of a distribution by naming only the mean and standard deviation, even when the graph is strongly skewed or has outliers.
Students memorize 'SOCS' and reach for the default summary statistics without checking whether the mean is an appropriate measure of center for the observed shape.
What to do instead
Before summarizing, classify the shape from the graph; if skewed or outliers are present, report median and IQR and explicitly justify the choice by referencing the shape.
Interpreting a standard deviation or z-score as a count or percentage of individuals rather than as a typical distance from the mean.
The formula is learned procedurally, so students never attach a units-based verbal meaning to the value.
What to do instead
Practice writing the full interpretive sentence template: 'The [context variable] typically differs from the mean of [x] by about [sd] [units].'
Concluding a study proves causation after a well-designed observational study, or generalizing a volunteer-sample result to the whole population.
Students conflate 'random' with 'random assignment' and treat any large sample as representative.
What to do instead
For every study, classify the design on two axes (random selection? random assignment?) and state the scope of inference permitted by each axis before writing any conclusion.
Computing a two-event probability by adding when the events are not disjoint, or multiplying when the events are not independent.
Students match a keyword ('or' to addition, 'and' to multiplication) to a rule without testing the disjointness or independence condition.
What to do instead
Before applying a rule, write out whether overlap or dependence exists; use the general addition rule P(A or B) = P(A) + P(B) - P(A and B) and check independence with P(A|B) = P(A).
Treating a sampling distribution of a sample proportion or mean as if it were the population distribution, including using the population standard deviation instead of the standard error.
Students do not distinguish the distribution of the variable from the distribution of the statistic across repeated samples.
What to do instead
For each problem, label the distribution as population, sample data, or sampling distribution, then use SE = sqrt(p(1-p)/n) or s/sqrt(n) only for the sampling distribution.
Checking conditions for a sampling distribution by writing 'np >= 10' without substituting numbers or stating the conclusion, and skipping the independence or 10% condition entirely.
Condition-checking is treated as a ritual checklist rather than evidence that a procedure is valid.
What to do instead
Write each condition as a sentence with substituted values and a verdict: 'np = 200(0.3) = 60 >= 10, so the sampling distribution is approximately normal; the sample of 200 is less than 10% of all students, so observations are independent.'
Interpreting a 95% confidence interval as 'there is a 95% probability the true parameter lies in this interval' or as capturing 95% of sample statistics.
The confidence level is attached to the wrong object because the repeated-sampling meaning is abstract.
What to do instead
Use the template: 'We are 95% confident that the true [parameter in context] is between [a] and [b],' and separately state the method captures the parameter in 95% of all possible samples.
Writing hypotheses with a sample statistic (e.g., H0: p-hat = 0.5) or stating a conclusion as 'we accept H0' or 'the probability H0 is true is 0.03.'
Students blur the sample and the parameter and treat the p-value as a posterior probability.
What to do instead
Always write hypotheses with population parameters; interpret the p-value as the probability of data this extreme assuming H0 is true; conclude 'reject H0' or 'fail to reject H0' in context.
Reporting a p-value or decision without linking it back to the original research question, or claiming practical significance from a small p-value in a huge sample.
Students treat the calculation as the endpoint and forget the context sentence is the graded conclusion.
What to do instead
End every inference with a two-part sentence: the statistical decision tied to alpha and a plain-language statement about the parameter in context; separately note whether the effect size is meaningful.
Interpreting a regression slope as a correlation, predicting far outside the observed x-range, or describing r-squared as the percent of points on the line.
Slope, correlation, and coefficient of determination are learned as interchangeable 'regression numbers' without unit-level meaning.
What to do instead
Write slope as 'predicted change in y-hat per one-unit increase in x,' r-squared as 'percent of variation in y explained by the linear model in x,' and restrict predictions to the observed range of x.