Common Mistakes in Statistics
The statistics mistakes students make most often are reading a correlation as proof of cause, using the mean on strongly skewed data where the median fits better, and confusing P(A | B) with P(B | A). Below are 20 common mistakes grouped by topic, each with how to fix it.
Mean vs Median
Using the mean automatically even when the data are strongly skewed
The mean is pulled by extreme values. When a distribution is skewed, the median often gives a better description of a typical value.
Thinking the median is just the “middle-looking” number without ordering the data
The median is defined after the values are put in order. If the data are not ordered, the chosen value is not the median.
Ignoring outliers before choosing a measure of center
One very large or very small value can change the mean a lot while barely moving the median.
Treating mean and median as interchangeable because both are called averages
They answer slightly different questions. The mean balances the data; the median splits the ordered data into two equal halves.
Choosing a center without checking the shape of the distribution
Always look at the graph or ordered list first. Shape, symmetry, and outliers tell you which summary is more trustworthy.
Sampling Bias
Treating a convenient sample as if it represents the whole population
Easy-to-reach people are not automatically representative. Convenience sampling often overrepresents one group and misses another.
Forgetting that who does not respond can matter as much as who does
Nonresponse can bias results if the people who skip the survey differ in an important way from those who answer it.
Assuming “large sample” means “unbiased sample”
A large biased sample is still biased. Sample size reduces random variability, not bad selection methods.
Generalizing from one class, school, or website to a much broader population
Check whether the sample frame matches the population you want to make claims about.
Ignoring wording or measurement choices that push responses in one direction
Bias can enter through the question itself, not only through who gets selected.
Correlation vs Causation
Assuming that if two variables move together, one must cause the other
Association alone does not identify cause. A third variable or a hidden process could explain the pattern.
Ignoring confounding variables
A confounder is related to both variables and can create the appearance of a direct causal link when none exists.
Using observational data to make a strong causal claim
Observational studies are useful, but strong causal claims usually require well-designed experiments with random assignment.
Reading a scatter plot as proof instead of evidence to investigate
A scatter plot can suggest direction and strength of association, but it does not establish mechanism.
Forgetting that causation usually needs a study design argument, not just a numerical one
Ask how the data were collected before deciding what kind of conclusion is justified.
Conditional Probability
Keeping the original denominator after the condition changes the sample space
In P(A | B), the condition B becomes the new universe. The denominator must describe the cases where B happens.
Confusing P(A | B) with P(B | A)
Reversing the condition changes the meaning. These probabilities can be very different even when the same two events are involved.
Assuming conditional probability means the events are independent
Independence is a special case where conditioning does not change the probability. Most conditional questions involve dependence.
Reading the wrong row or column in a two-way table
Mark the conditioned group first, then compute the proportion within that group only.
Using multiplication rules without checking whether the events depend on each other
If the first event changes the chance of the second, you need a conditional probability in the multiplication rule.
Mistakes From Every Lesson, by Topic
Every statistics lesson lists the mistakes students make with that idea. These pages collect them by topic, each linked to the lesson that explains the right idea.
- Center, Spread & Distributions (76 mistakes)
- Data Collection & Displays (64 mistakes)
- Probability & Chance (52 mistakes)
- Relationships & Regression (44 mistakes)
- Sampling, Design & Inference (76 mistakes)
Why These Mistakes Happen
Statistics mistakes usually happen when a summary, graph or probability is asked to support a claim it cannot support. The calculation can be right and the conclusion still wrong, so most fixes below are about what a result allows you to say.