Start with the course you actually have
Statistics is not one fixed sequence of topics. A course may emphasise probability, data analysis, inference, regression or a mixture of these. Begin with your syllabus, lecture notes and assessment guidance. List the topics and the methods you are expected to use; do not assume that a textbook’s chapter order matches your course. If you are preparing for a formal examination, check the relevant institution’s official guidance for its structure and permitted materials.
For each topic, collect the material you already have: lecture slides, worked examples, problem sheets, textbook notes and any handwritten pages. Make a simple index of the topic names and where the relevant material is stored. This avoids spending study time searching and shows you which areas have no usable notes yet.
Then sort each topic into three practical questions:
- What is the idea? For example, what does a confidence interval describe?
- How do I choose the method? For example, when is a paired comparison appropriate?
- How do I carry it out and interpret it? Include assumptions, calculations and a conclusion in context.
A revision board should make these connections visible rather than collect formulas without a reason to use them. A board on this topic ends up looking like this:
- For independent observations with finite variance, the sample mean has standard error σ/√n.
- For a sample proportion, estimated SE = √[p̂(1−p̂)/n]; check that the normal approximation is reasonable.
- A standard error describes sampling variability, not the spread of individual observations.
| Quantity | Interpretation |
|---|---|
| Slope b₁ | Predicted change in response for one-unit increase in predictor |
| Residual | Observed response minus fitted response |
| R² | Proportion of observed response variation accounted for by the fitted model |
Estimate ± critical value × standard error. State the parameter and interpret the interval in context; a 95% procedure has 95% long-run coverage under its assumptions.
- Inspect residuals against fitted values for curvature and changing variance.
- Check independence from the study design; a plot alone cannot establish it.
- Investigate influential observations and avoid extrapolating beyond the data range.
- Set H₀ and H₁ before calculating; choose a one- or two-sided alternative from the question.
- A p-value is the probability, assuming H₀ and the model, of data at least as extreme as observed.
- Reject H₀ at level α when p ≤ α; this does not establish that H₀ is false.
The board is a working map, not a replacement for your course material. Compare it with your syllabus and add course-specific notation, conditions and examples. If your lecturer uses a different convention, record that convention and use it consistently in assessed work.
Build a method-selection habit
Many lost marks in statistics come before the arithmetic: a student uses a method that does not answer the question, or states a result without its conditions. For every worked problem, pause before calculating and write a short method plan:
- Identify the observational unit, response variable and any explanatory or grouping variables.
- Decide what the question asks for: description, estimate, comparison, association or prediction.
- Name the method and state why it fits the data and design.
- List assumptions or conditions that need checking.
- Calculate, then interpret the result using the variables in the question.
For instance, a comparison of measurements before and after an intervention on the same people is paired: analyse the within-person differences rather than treating the two sets of measurements as independent. A comparison between separate groups calls for a method suited to independent samples. The exact test and conditions depend on the data and the course’s treatment of the topic.
Practise this plan with questions whose surface details vary. A question about plant growth and one about test scores may use the same statistical structure. Covering the method name in your notes and deciding from the design helps you recognise that structure instead of memorising a trigger word.
Learn a small set of core statements precisely
Create a short core checklist from your own notes. Include definitions that are easy to blur, such as standard deviation versus standard error, and interpretations that require careful wording. Keep formula conditions beside the formulas: an expression used outside its assumptions is not a reliable method.
A useful checklist for an introductory inference and regression unit could look like this:
Test yourself by explaining each statement without looking, then check against your notes. Where a statement is incomplete, correct it rather than merely rereading it. For example, “95% confident” is only useful when you can explain the long-run coverage meaning and identify the assumptions behind the interval.
Practise recall, then solve full problems
Use brief question-and-answer cards for definitions, interpretations and method conditions. Keep one card to one fact or distinction. Cards are good for retrieving a formula or explaining a p-value; they are not a substitute for working through a multi-step problem, checking assumptions or writing a conclusion.
A statistics card set should prompt an answer rather than display a paragraph to memorise. A card drill on this board looks like this:
What does the standard error of a sample mean measure?
The estimated or known standard deviation of the sampling distribution of the sample mean.
All 7 cards
That is a MySummaries deck, filled with statistics material. Yours is written from your own notes. Start free
Work through each card by saying or writing the answer before revealing it. Grade honestly: an answer that remembers a phrase but misses the meaning needs another review. For numerical topics, make calculation cards only when the calculation is short and has a clear target. Keep longer procedures in problem practice, where you can show each step and check units, rounding and interpretation.
After recall, solve a small set of problems from scratch. Start with one example where the method is familiar, then vary the sample size, context or wording. For each answer, check four things: the method matches the design, the calculation is correct, the assumptions have been considered, and the final statement answers the question. Keep an error log with the cause of each mistake—for example, “treated paired observations as independent”—and add a new practice question that targets that cause.
Use worked solutions without becoming dependent on them
When you consult a worked solution, compare its reasoning with yours line by line. Do not only compare the final number. Ask where the solution identifies the parameter, chooses a distribution or test, checks conditions and interprets the result. If you cannot explain a step, return to the relevant section of your notes and write a short explanation in your own words.
For computation, use the tools allowed by your course and practise with them before an assessment. Check what the output represents: software may report a standard error, test statistic, confidence interval or p-value, but it cannot decide whether the method is justified by the design. Preserve enough working to show how the output answers the question.
Once a week, complete a mixed set without arranging questions by topic. This tests whether you can identify a method when it is not announced in advance. Review errors straight afterwards, while you can still see what decision led to them. Spend the next session on the weakest recurring issue, not simply on the most recently taught topic.
Listen to a concept explained aloud
Audio can be useful for reviewing a connected idea while walking or travelling, but it should not replace active problem-solving. Use it to rehearse explanations, then pause and reproduce the reasoning or solve a related question yourself. A short lecture made from the board might sound like this:
Transcript · tap any word to jump there
Start with the null hypothesis and the model used for the test. The p-value is calculated as if those assumptions hold. It measures how unusual the observed result, or a more extreme result in the direction defined by the test, would be under that setup.
A small p-value is evidence against the null model at the chosen significance level. It is not the probability that the null hypothesis is true, and it does not measure the size or practical importance of an effect. A large p-value is not proof of no effect; the data may simply be imprecise.
Finish by reporting the result in context. State the evidence, relate it to the question and note relevant design or model limitations. Do not turn statistical significance into a causal claim unless the study design supports one.
A repeatable weekly routine
A practical week can have three or four sessions, adjusted to your timetable. In a 45-minute session, spend about 10 minutes on due recall, 25 minutes solving problems and 10 minutes correcting errors. In a longer session, reserve a final block for a mixed question or a concise explanation of a concept. The important part is the sequence: retrieve, apply, inspect and repair.
At the end of each week, update your board from the errors you actually made. Mark topics that need another problem, revise definitions that remain uncertain and remove cards you can consistently answer. Keep a short list of unresolved questions to take to a tutor, lecturer or study group. This makes your next study session specific rather than a general rereading session.
How MySummaries helps
MySummaries lets you build a statistics revision board from your own PDFs, slides and photographed notes, then use that board to make recall cards and examiner-voiced audio lectures. For this method, use the board to connect each course topic to its conditions and examples, review cards for precise definitions, and listen to a concept before testing it with your own problem set.