Statistics for Beginners

Begin statistics with the habit of asking what a number describes. Learn to distinguish a population from a sample, inspect a distribution, calculate simple summaries, and explain uncertainty. This roadmap uses a fictional waiting-time study rather than advanced formulas. Your final output is a short analysis that another learner can challenge and reproduce. The purpose is careful interpretation, not turning a tiny exercise into a confident claim about everyone.

Define what was observed

Imagine a fictional help desk recording the waiting time of five visitors during one morning. The times are 2, 3, 3, 4, and 18 minutes. Write the unit, location, period, and selection method next to the data. The observational unit is one visitor’s wait. The five visitors form the observed sample; they are not automatically representative of every day, shift, or visitor.

Before calculating anything, ask how the measurements were made. Did waiting start on arrival or after registration? Were people who left excluded? Did one observer choose only easy cases? These questions can matter more than using an advanced statistical technique. OpenIntro offers a structured introductory statistics text for further study Source 1; use it to deepen concepts after examining your own small example.

Describe centre and spread together

The five waits total 30 minutes, so the mean is 6 minutes. The middle sorted value is 3 minutes, so the median is 3. The range is 18 minus 2, or 16 minutes. Each number captures something different. A reader told only that the typical wait is six minutes may miss both the cluster of short waits and the unusually long one.

Draw five dots on a number line and label their values. Keep repeated values visible rather than merging the two three-minute waits into one observation. Statistics Canada explains range and interquartile range as measures of spread Source 2. You can learn quartiles next, but first make sure you can explain why the 18-minute observation changes the mean more strongly than the median.

Do not delete the long wait merely because it is inconvenient. Check whether it is a recording error, a different kind of visit, or a genuine event. In this fictional exercise it is genuine. Report the full dataset and describe what further information would help explain it.

Separate description from an explanation

A summary describes the recorded waits. It does not show why a person waited 18 minutes. Perhaps the request was complex; perhaps staff were temporarily unavailable; perhaps an earlier queue caused a delay. These are hypotheses, not findings. Make a two-column note separating observations from possible explanations, then propose what additional data would discriminate among them.

Suppose you observe five waits on a second fictional morning: 2, 2, 3, 4, and 4 minutes. Their mean is 3 and median is 3. You can say the recorded average was lower on that morning. You cannot attribute the change to a new queue sign unless the study design addresses other differences. NIST’s engineering statistics handbook provides background on exploratory analysis and experimental design Source 3.

Learn probability and confidence intervals only after this distinction is clear. An interval is not a remedy for a poorly defined sample. A precise calculation based on biased observations can still answer the wrong question. Write the limitations before you see the result so they do not change according to whether the result is convenient.

Create a small statistical memo

Prepare a memo containing the question, raw values, dot plot, mean, median, range, and two limitations. State that the data is fictional and too small to represent a service-wide performance estimate. Recommend a next collection plan: consistent timing rules, several periods, and a record of visit type without unnecessary personal information.

For a review exercise, change 18 to 8 and recompute. The total becomes 20, the mean 4, the median remains 3, and the range becomes 6. Explain which summaries changed and why. Then restore the original data. Keep both versions labelled so a reviewer does not mistake a sensitivity exercise for a corrected observation.

Finish by asking a peer to identify one claim your data supports and one it does not. If they infer causation from the two mornings, strengthen your explanation. Progress to larger datasets and formal inference when you can already communicate a simple sample honestly.

Practice sequence

DefineSpecify the unit, sample, and timing rule.
DisplayShow individual values before summarising.
CalculatePair a centre measure with spread.
QualifyState what the observations cannot establish.

Practice decisions and evidence
Checkpoint	What to produce	How to review it
Centre	Mean 6; median 3 minutes	Compute from all five original waits
Spread	Range 16 minutes	18 minus 2, with units retained
Sensitivity	Mean becomes 4 if 18 becomes 8	Median remains 3
Interpretation	One fictional morning	No population or causal claim

Evidence before advancement

Observation: One visitor waited 18 minutes.
Hypothesis: Their request may have been more complex.
Needed evidence: Visit type and comparable timing records.

Common questions

Do I need advanced mathematics to start?

You can begin this roadmap with arithmetic, fractions, and a willingness to reason carefully about data. Learn additional mathematics when a concept requires it. Being able to explain the sample and denominator is immediately useful. Memorising a formula without knowing what its inputs represent will not make an interpretation reliable.

Which average should I report?

Choose the summary that suits the reader’s question and show enough context to prevent misunderstanding. In the waiting-time example, reporting mean and median together reveals the influence of the long wait. Avoid presenting one as universally correct. The distribution and purpose determine which summary is most informative for the task.

Should an outlier always be removed?

No. Investigate whether the unusual value is erroneous, belongs to a different process, or is a valid observation. Record your decision and its reason. In the exercise the long wait is genuine, so it stays. Removing it solely to produce a nicer average would conceal information the reader may need.

Does a larger sample automatically remove bias?

A larger sample can still be systematically unrepresentative. Recording only quiet mornings, for example, may miss busy periods no matter how many quiet-morning visitors you measure. Examine selection and measurement rules as well as sample size. Better coverage and consistent definitions are essential parts of a stronger collection plan.

Can I say the new queue sign reduced waiting?

Not from the two fictional mornings alone. The observed difference could reflect visit types, staffing, demand, or chance. You need a design that addresses alternative explanations before making a causal claim. For now describe the recorded difference and propose a more careful comparison rather than giving the sign credit automatically.

What is the difference between a sample and a population?

The population is the group your question concerns; the sample is the portion actually observed. If your question concerns all help-desk visits during a month, five visits on one morning are a limited sample. State both clearly so readers can judge how far the findings might reasonably extend.

When should I learn hypothesis testing?

Learn it after understanding distributions, sampling, probability, and the difference between a question specified in advance and one noticed afterwards. Formal testing adds assumptions and interpretation choices. Start by describing small datasets accurately. A test result does not replace subject knowledge or repair unclear measurement and selection procedures.

How do I show statistical reasoning in a portfolio?

Include the raw values, calculations, visualisation, sampling description, and limitations. Show the sensitivity exercise and explain why it is not a data correction. A short memo that separates observation from speculation demonstrates more judgement than a page of unexplained formulas or a claim of certainty based on a tiny sample.