Understand the idea behind the answer.
Establishing a statistical foundation, choosing an appropriate approach for collecting or analysing data, and interpreting results with their uncertainties are key skills that statistical calculators can help develop.
Learn by topic
17 topicsDescriptive statistics
Describe your dataset prior to attempting to explain it. Compare your datasets' center and variability, check out the outliers, and determine what summary statistic(s) most clearly address your research question.
Explore topicProbability
Keep track of which outcomes could occur and how new information changes their chances. Begin with simple events. Then use conditional probability and Bayesian updates.
Explore topicNormal distribution
A mean determines the center point of the normal distribution, while a standard deviation determines the spread. Normality provides a symmetric distribution which is beneficial for certain types of measurement or sampling distributions. Start by determining if the distribution of your data is consistent with your questions; having a mean and standard deviation does not automatically make your data normally distributed.
Explore topicBinomial distribution
Count successful observations over a fixed number of independent trials. The model has to have one single success rate; if there is changing probability or dependency among trials, the calculation will also change accordingly.
Explore topicSampling distributions
Sample statistics vary between samples. Standard errors describe that sampling variability under a model; they do not describe variation among individual observations.
Explore topicConfidence intervals
Report an estimate with its uncertainty. The interval method depends on the parameter, sampling design and assumptions. Confidence describes the method’s long-run coverage when it is used repeatedly.
Explore topicHypothesis testing
Before looking at your test results, identify the model and alternative. Interpret the p-value alongside the effect size, uncertainty and quality of the study.
Explore topicOne-sample t-test
Use one sample to compare a population mean with a specified reference. The t calculation accounts for estimating the population standard deviation from those same observations.
Explore topicTwo-sample t-test
Compare two independent group means while accounting for uncertainty in both. Welch’s method does not require equal population variances; a pooled t calculation makes that additional assumption.
Explore topicPaired samples t-test
Compare differences within meaningful pairs, such as repeated measurements on the same person. The paired t-test takes the difference within each pair, then applies a one-sample test to those differences.
Explore topicSimple linear regression
Draw a straight line relating one variable that is numeric to another that is also numeric. Review the distribution of points (scatter) and the residuals before making use of this line in a predictive role or drawing conclusions from its slope.
Explore topicPearson correlation
Measure the direction and strength of a linear association. Inspect the observations: extreme values or a nonlinear pattern can make a single coefficient misleading.
Explore topicData visualisation
Pick an illustration that keeps intact the comparisons you want to compare. Illustrations should include units, scales and sufficient individual data values to show variability behind some measure of central tendency.
Explore topicStudy design
Before beginning to collect data define your population, outcome(s), and comparison. The way in which you randomly assign participants to either the treatment group or the control group along with sampling and follow-up represent ways of dealing with the various forms of uncertainty and bias associated with research.
Explore topicInterpreting research
Determine the answer to these three questions regarding an article: What were they estimating? How much uncertainty exists in their estimate? Do the design of the study support the authors' conclusions? Statistical evidence must be interpreted within a given context and by itself a "threshold" will not provide answers to all three of these questions.
Explore topicChecking assumptions
Before using a calculator, check what its model assumes. Observations may need to be independent, and the measurements must represent the quantity of interest. The required distributional assumptions depend on the method.
Explore topicBayesian reasoning
Combine a prior distribution with the likelihood of observing the given data. Be explicit about both parts of this input combination, and investigate how variations in other reasonable inputs will influence the resulting posterior distribution.
Explore topicWorked guides
12 concise explanationsMean, median and mode: choosing a useful average
There are three common measures of a typical value. An example illustrates why a particular measure might be preferable to another.
Read the explanationFoundations · 3 min readStandard deviation: measuring how much values vary
Understand the concept of variability, differentiate between sample and population measures, and maintain a distinction between standard deviations and standard errors.
Read the explanationInference · 3 min readConfidence intervals: putting uncertainty around an estimate
Know how to read an interval correctly and know why sample size, variability and confidence level change width of interval.
Read the explanationInference · 3 min readP-values: what they answer and what they leave out
A p-value shows how compatible your data are with the statistical model you chose. It doesn’t tell you the probability that your hypothesis is correct.
Read the explanationInference · 3 min readChoosing a statistical test from the question and study design
Start with the outcome and comparison. Ask how the observations are related, including whether they are paired. Those choices help determine the method.
Read the explanationInference · 3 min readOne-sample, paired and Welch t-tests
Match the t-test to your comparison. Then see how the estimated mean or difference is divided by its standard error to produce the test statistic.
Read the explanationResearch · 3 min readCorrelation, association and claims about cause
Correlation describes a relationship or pattern; however, we must keep the calculation of this pattern separate from the causal explanation of why the relationship exists.
Read the explanationInference · 3 min readSample size for estimating a proportion
Consider planning a basic proportion estimate for the precision you desire. When selecting your option for calculations, distinguish study power from survey precision.
Read the explanationResearch · 3 min readAbsolute risk, relative risk and number needed to treat
Translate a relative effect into actual event counts, showing the baseline event risk and the follow-up period.
Read the explanationResearch · 3 min readSensitivity, specificity and the meaning of a positive test
Examine test performance and predictive value using a full two-by-two table.
Read the explanationResearch · 3 min readReading a research paper beyond its headline
When reviewing an article begin with the question asked, then review the design used to obtain the data upon which the findings are based. Followed by the point estimate and measures of precision prior to determining if you agree or disagree with the author's conclusion.
Read the explanationFoundations · 3 min readProbability distributions: matching a model to the question
Relate possible outcomes, probability models and sampling distributions without assuming that all datasets are normally distributed.
Read the explanation