Classroom · Statistics and data science
Biostatistics
Distributions, tests and study design.
133 lessons, text and equations; animations coming soon.
Early edition: these lessons are being reviewed.
Introductory level (39)
- Central limit theoremThe central limit theorem says that the average of many independent observations tends to follow a normal distribution, whatever the shape of the original data…
- Confidence interval coverageA 95% confidence interval is produced by a procedure that, over many repeated studies, captures the true population parameter about 95% of the time.
- Two-sample t-testThe two-sample t-test asks whether the means of two independent groups differ by more than random sampling variation would explain.
- P-value under the nullA p-value is the probability, assuming the null hypothesis is true, of getting data at least as extreme as what we observed.
- Card ShufflingThis lesson models card shuffling using the riffle shuffle, where a deck is cut and interleaved, linking the process to rising sequences and a-shuffles.
- Using RR is a free, open-source statistical computing environment where users enter commands at a prompt, typically using GUIs like RStudio.
- Data Science in PracticeData science bridges technical analysis and business intelligence through effective reporting and visualization, enabling non-technical stakeholders to make informed…
- Data and DatasetsData, the plural of datum, refers to any analyzable information, while a dataset organizes these observations into items (rows) and attributes (columns).
- Rectangular DataRectangular data is the foundational two-dimensional matrix structure in data science, where rows represent records (cases/observations) and columns represent features…
- Data Science with PythonThis lesson introduces the Python data science workflow using Jupyter Notebooks in Google Colab, focusing on the Pandas library for data manipulation and Matplotlib for…
- Exploring Binary and Categorical DataCategorical data is summarized using proportions, percentages, and the mode, which identifies the most frequent category.
- Exploratory analysis of Galton's DataThis lesson explores Galton's 1885 anthropometric data, highlighting the 1.08 scaling factor applied to female heights to adjust for gender differences.
- Prevalence and IncidenceProbability theory provides the essential framework for distinguishing statistical significance from random chance in research.
- The Basic ElementsA frequency distribution organizes data into a tabular or graphical summary showing how often specific values occur.
- Measures of the Location of the DataMeasures of location, including the mean, median, mode, quartiles, and percentiles, summarize data central tendency to aid population inference.
- Stem-and-Leaf Graphs (Stemplots), Line Graphs, and Bar GraphsStem-and-leaf plots preserve exact data values while revealing distribution shapes, making them ideal for small datasets.
- Box PlotsA box plot visualizes data concentration and extremes using the five-number summary: minimum, Q1, median, Q3, and maximum.
- Measures of the Spread of the DataMeasures of dispersion quantify how data points vary around a central value, complementing central tendency metrics.
- The varianceThe Central Limit Theorem states that for large sample sizes, the sampling distribution of the mean approximates a normal distribution with mean μ and standard error…
- Histograms, Frequency Polygons, and Time Series GraphsHistograms and frequency polygons are graphical tools used to visualize the shape, center, and spread of large datasets, typically containing 100 or more observations.
- The dataThis lesson introduces the fundamental categorization of data as either ungrouped or grouped and defines measures of central tendency, including the mean, median, and…
- Data, Sampling, and Variation in Data and SamplingData is classified as qualitative (categorical) or quantitative (numerical), with the latter subdivided into discrete and continuous types based on whether values are…
- Frequency, Frequency Tables, and Levels of MeasurementFrequency tables organize data by counting occurrences, where the sum of all frequencies equals the sample size n.
- Experimental Design and EthicsExperimental design investigates causal relationships by manipulating an explanatory variable and measuring a response variable, using random assignment to control for…
- Data Collection ExperimentData collection is the systematic gathering of information from a sample or population to enable sound interpretations.
- Some Other Things You Should KnowThis lesson clarifies the distinction between sample statistics and population parameters, emphasizing the need for rigorous notation to avoid confusion.
- Univariate DataUnivariate analysis examines a single variable's distribution, central tendency, and spread, requiring rigorous probability sampling and a minimum sample size of 30.
- TerminologyThis lesson defines the foundational elements of probability, including experiments, trials, outcomes, and sample spaces.
- Probability TopicsThis lesson introduces the visualization of events using Venn diagrams and the addition rule for calculating the probability of unions, illustrated by a blood type…
- Tree and Venn DiagramsVenn diagrams and tree diagrams are visual tools that translate abstract probability concepts into geometric representations.
- Independent and Mutually Exclusive EventsIndependence and mutual exclusivity are distinct probabilistic relationships; independent events do not influence each other's likelihood, while mutually exclusive…
- Permutations and CombinationsPermutations and combinations are foundational tools in biostatistics for calculating discrete probabilities, distinguished by whether the order of selection matters.
- PercentagesPercentages standardize data by expressing a proportion per hundred, facilitating comparison across different group sizes.
- The dice problemThe expected value E(X) is the weighted average of outcomes, serving as the long-run mean and center of mass for a random variable.
- Collection and Compilation of DataThis lesson covers the distinction between population variation and estimate precision via the standard error, the computation of Hedges' g for meta-analysis, and the…
- Oliver's bloodOliver's blood problem demonstrates that data consistent with a hypothesis does not necessarily provide evidence in its favor.
- Risk ratioThe risk ratio, or relative risk, is a measure used in cohort studies to compare the probability of an event occurring in an exposed group versus an unexposed control…
- Research Questions About One GroupThis lesson covers methods for comparing two independent groups, focusing on the two-sample t-test for numerical means and tests like chi-square or Fisher's exact for…
- One-Tailed and Two-Tailed Test ScenariosOne-tailed and two-tailed tests are selected based on whether the alternative hypothesis is directional or non-directional.
Intermediate level (87)
- Statistical power and effect sizeStatistical power is the probability that a study detects a true effect of a given size, equal to 1 minus the Type II error rate.
- Bonferroni and FDR controlWhen many hypothesis tests are run, the chance of at least one false positive grows quickly.
- Bootstrap resamplingBootstrap resampling estimates the uncertainty of a statistic by repeatedly drawing samples with replacement from the observed data.
- Permutation testA permutation test asks whether an observed difference between groups could plausibly arise by chance if group labels were meaningless.
- Kaplan-Meier estimatorThe Kaplan-Meier estimator estimates the survival function from time-to-event data in which some patients are censored.
- ANOVA variance partitionOne-way ANOVA splits the total variation in a measurement into variation between group means and variation within groups.
- Bayesian prior to posteriorBayesian updating combines what we believed before seeing data (the prior) with how well each hypothesis explains the data (the likelihood) to give an updated belief…
- ROC curve and AUCA ROC curve plots sensitivity against 1 minus specificity as the cutoff of a diagnostic test is varied.
- Maximum likelihood estimationMaximum likelihood estimation (MLE) chooses the parameter value that makes the observed data most probable under a assumed model.
- Row Echelon FormRow echelon form (REF) and reduced row echelon form (RREF) are structural matrix states defined by specific pivot and zero-row criteria.
- Linear MapsLinear maps are functions between vector spaces that preserve vector addition and scalar multiplication, forming the structural backbone of linear algebra.
- Case Study: Effects of Tobacco Use on Bone-Mineral Density (BMD) in Middle-Aged WomenThis lesson examines a 1994 study by Hopper and Seeman that used a matched-pair design of 41 female twin pairs to isolate the effect of tobacco use on bone-mineral…
- The sample varianceSample variance (S²) quantifies data dispersion by averaging the squared deviations from the sample mean, using n−1 in the denominator to correct for bias.
- Case Study: Effects of Lead Exposure on Neurologic and Psychological Function in ChildrenThis lesson analyzes a biostatistical study on the effects of lead exposure on children's neurological and psychological function, utilizing data from El Paso, Texas.
- Tests of SignificanceBiostatistics provides the methodology to distinguish true effects from natural variation using statistical significance, typically defined by a P-value less than 0.05.
- Exploring Two or More VariablesThis lesson covers bivariate and multivariate analysis techniques for numeric and categorical data, emphasizing visualization methods like hexagonal binning and…
- How a Meta-Analysis WorksA meta-analysis synthesizes individual study results by calculating an effect size for each study, which serves as the fundamental unit of analysis.
- PrivacyData privacy is defined as the methodological balance between preserving data utility for statistical analysis and mitigating the risk of re-identifying individuals.
- The goals of inferenceStatistical inference bridges sample data and population truths through five core goals: estimating quantities with uncertainty, hypothesis testing, inferring…
- Confidence Interval and Confidence LevelDiagnostic accuracy relies on sensitivity (true positive rate) and specificity (true negative rate), as demonstrated by FNAC's 95% and 93% performance respectively.
- Other Considerations in the Regression ModelThis lesson covers advanced regression considerations, distinguishing between confidence and prediction intervals, and handling qualitative predictors via dummy…
- Additional Topics and ApplicationsThis lesson covers the adjoint matrix, Cramer’s Rule for solving linear systems, and the cross product in vector calculus.
- The M&M problemThis lesson applies Bayesian inference to the M&M problem, demonstrating how to update beliefs about bag origins using color distributions.
- Survey Design and ImplementationSurveys gather data from a sample to infer characteristics of a larger population, requiring careful design to minimize bias and error.
- Web Scraping and Social Media Data CollectionThis lesson covers automated data acquisition via web scraping and social media APIs, highlighting Python libraries like pandas and re for extraction and processing.
- Data Cleaning and PreprocessingData cleaning and preprocessing transform raw information into accurate, consistent structures through six sequential steps: integration, cleaning, transformation…
- Multi-Arm Bandit AlgorithmMulti-arm bandit algorithms offer a dynamic alternative to static A/B testing by continuously optimizing for cumulative rewards through an exploration-exploitation…
- Vector SpacesVector spaces are mathematical systems defined by addition and scalar multiplication that obey specific algebraic rules, generalizing concepts from Euclidean spaces…
- ParadoxesProbabilistic paradoxes stem from ambiguous information acquisition rather than mathematical errors, requiring explicit definition of sampling mechanisms.
- Bayes's theoremBayes's theorem provides a mathematical framework to reverse conditional probabilities, allowing clinicians to calculate the probability of a disease given a test…
- IID random variablesIID random variables are defined by mutual independence and identical population origin, forming the mathematical ideal for random samples.
- The odds form of Bayes's theoremThe odds form of Bayes's theorem simplifies belief updating by expressing the relationship between prior and posterior probabilities as a ratio, effectively eliminating…
- Contingency TablesContingency tables organize categorical data into R rows and C columns to analyze relationships between variables.
- Effect Sizes Based on CorrelationsWhen analyzing correlations between continuous variables, the correlation coefficient r serves as the effect size index, but its variance depends strongly on the…
- The Variance of a Discrete Random VariableThis lesson defines the variance of a discrete random variable as a measure of spread relative to the expected value, calculated using the formula…
- Mean-variance relationshipThe Poisson distribution is defined by a single parameter where the mean and variance are identical.
- Beyond the Basics: Comparisons and ConclusionsThis lesson explains how distribution curves visually represent data, where area corresponds to observation counts and width reflects standard deviation.
- The Homogeneity AttackThe homogeneity attack exploits k-anonymity's failure to ensure diversity in sensitive attributes, allowing attackers to infer private information when all records in…
- Finite-Dimensional Vector SpacesFinite-dimensional vector spaces are defined by the existence of a finite spanning list, enabling coordinate systems and matrix representations.
- An alternative priorArbitrary uniform priors can cause high posterior sensitivity when data is scarce, but this effect diminishes as more data is collected.
- Factors that Affect PrecisionPrecision in meta-analysis is defined by variance, standard error, and confidence intervals, which are mathematically linked and affected by sample size and study design.
- SubspacesA subspace is a nonempty subset of a vector space that is closed under vector addition and scalar multiplication, inheriting all vector space axioms from the parent…
- A Special CurveThe normal distribution is a symmetrical, unimodal curve defined exclusively by its mean and standard deviation, where the mean, median, and mode coincide at the peak.
- Regression to the meanRegression to the mean is a statistical phenomenon where extreme initial measurements tend to be less extreme in subsequent paired observations, a concept introduced by…
- Change of BasisChange of basis establishes a unique coordinate representation for vectors relative to an ordered basis, where the order of basis vectors strictly determines the order…
- Row Space and Column SpaceThe row space and column space of a matrix A are subspaces spanned by its rows and columns, respectively, with the rank defined as their shared dimension.
- From Data to Action - Executing Healthcare Analytics for ImpactHealthcare analytics success depends on translating raw data into actionable wisdom through structured frameworks like the DIKW hierarchy and the five-phase…
- Electronic Health Records and Data Infrastructure—The Foundation for Scalable Healthcare AnalyticsElectronic Health Records (EHRs) are enterprise-wide digital platforms that enable secure, real-time data exchange and interoperability across multiple care settings…
- Navigating the Waters of Healthcare Data: Warehouses and Lakes for Enhanced AnalyticsData warehouses and lakes serve complementary roles in healthcare analytics, with warehouses optimizing for structured reporting through schema-on-write and lakes…
- Diverse Perspectives and Advanced Techniques in Healthcare Data AnalysisHealthcare analytics has evolved from retrospective statistical evaluation to modern AI-driven approaches that integrate clinical, operational, financial, and…
- Balancing the Triangle: Data Insights into Payers, Providers, and PatientsThe Healthcare Iron Triangle models the interdependent relationship among patients, providers, and payers, governed by competing objectives of access, quality, and cost.
- Mapping the Terrain: Domains and Maturity in Healthcare AnalyticsHealthcare analytics integrates ten distinct data domains, including clinical, revenue cycle, and operational data, to provide holistic insights into patient care and…
- Probability Distribution Function (PDF) for a Discrete Random VariableA discrete probability distribution function (PDF) maps countable outcomes to their probabilities, requiring each probability to be between 0 and 1 and the total sum to…
- Geometric Distribution (Optional)The geometric distribution models the number of independent Bernoulli trials required to achieve the first success, characterized by a constant success probability p…
- The sample meanThe sample mean is the primary estimator for the population mean, calculated as the sum of observations divided by sample size.
- Binomial trialsThe binomial distribution models the count of successes in n independent Bernoulli trials with constant probability p.
- Poisson distributionThe Poisson distribution is a discrete probability model for unbounded count data within a fixed interval, defined by a mean parameter μ where variance equals the mean.
- Conditional distributionsConditional distributions allow researchers to estimate the probability of one variable given another by slicing joint distributions, enabling inverse probability tasks…
- Sequential CompositionSequential composition establishes that the total privacy cost for multiple mechanisms applied to the same dataset is bounded by the sum of their individual epsilon…
- Linear TransformationsA linear transformation is a mapping between vector spaces that preserves linear combinations, satisfying both additivity and homogeneity.
- Odds RatioThe odds ratio is a quantitative metric used to estimate the strength of association between an exposure and an outcome, particularly in case-control studies.
- Estimation for the Poisson DistributionPoisson estimation analyzes rare events by modeling the count X as a Poisson variable with mean μ equal to λ×T, where λ is the incidence rate and T is person-years.
- The Standard Normal DistributionThe standard normal distribution, denoted Z ~ N(0, 1), is a symmetric bell curve with a mean of 0 and standard deviation of 1.
- Normal Distribution (Pinkie Length)This lesson guides students through a statistical lab to determine if adult pinkie finger lengths follow a normal distribution by comparing empirical data against a…
- Using the Normal DistributionThe normal distribution is a symmetrical probability density function where the mean, median, and mode coincide, allowing probabilities to be calculated as areas under…
- Estimation for the Binomial DistributionBinomial estimation infers the population proportion p using the sample proportion p-hat, which is an unbiased estimator.
- Normal Distribution—Lap TimesThe normal distribution models continuous data like racing lap times using a symmetric, bell-shaped curve defined by the mean and standard deviation.
- SimilaritySimilarity describes how different matrix representations of the same linear operator are related when the basis of the vector space changes.
- Probability density functionsA Probability Density Function (PDF) maps continuous values to probability densities, where probability is defined as the area under the curve.
- The Exponential Distribution (Optional)The exponential distribution models the continuous time elapsed until a specific event occurs, characterized by a decay parameter m which is the reciprocal of the mean…
- The Uniform DistributionThe uniform distribution models scenarios where outcomes are equally likely, distinguishing between discrete finite sets and continuous intervals.
- Continuous DistributionContinuous random variables assume an infinite number of values within a range, requiring probabilities to be calculated as the area under the probability density…
- Binomial Distribution (Optional)The binomial distribution models the count of successes in a fixed number of independent Bernoulli trials, denoted as X∼B(n,p).
- UpdateBayesian updating combines historical prior data with a contestant's specific guess to form a posterior distribution, which reflects updated beliefs about the true price.
- Identifying and Quantifying HeterogeneityHeterogeneity in meta-analysis refers to the variation in true effect sizes across studies, distinct from random within-study error.
- Ordinary least squaresOrdinary least squares (OLS) is a statistical procedure that estimates regression parameters by minimizing the sum of squared differences between observed and predicted…
- The Scalar Product in RnThe scalar product of two vectors in Rn is the sum of the products of their corresponding components, resulting in a real number that geometrically relates to the…
- Inner Product SpacesAn inner product space generalizes the dot product to abstract vector spaces by assigning a scalar to each pair of vectors, preserving geometric concepts like length…
- Orthonormal SetsOrthonormal sets combine orthogonality and unit length, enabling simplified coordinate calculation via inner products and norm preservation through Parseval’s Formula.
- Generalized Linear ModelsGeneralized Linear Models (GLMs) extend linear regression to handle non-normal responses by combining an exponential family distribution, a linear predictor, and a link…
- Discriminant AnalysisDiscriminant Analysis, introduced by R. A. Fisher in 1936, is a statistical classifier that uses discriminant functions to maximize class separation. It relies on the…
- Evaluating Classification ModelsThis lesson introduces the confusion matrix and key metrics like precision, recall, and specificity for evaluating binary classification models.
- Avoiding Sensitivity UnderestimationSensitivity dictates the noise magnitude for differential privacy; underestimating it breaks privacy guarantees while overestimating it reduces accuracy.
- Sums of Continuous Random VariablesThe distribution of the sum of independent continuous random variables is determined by the convolution of their individual probability density functions.
- Introduction to Time Series AnalysisTime series analysis involves examining numerical data points collected at specific time intervals to identify trends, seasonality, and noise.
- Time Series Forecasting MethodsTime series forecasting relies on decomposing data into trend, seasonal, cyclic, and residual components to build predictive models.
- The Wilcoxon Rank-Sum TestThe Wilcoxon Rank-Sum Test is a nonparametric alternative to the independent-samples t-test used when data is not normally distributed or is ordinal.
Advanced level (7)
- Mixed model shrinkageIn a linear mixed model, group-specific effects are estimated as a weighted compromise between each group's own data and the overall population mean.
- Metropolis MCMC samplerThe Metropolis algorithm draws samples from a probability distribution known only up to a normalising constant, such as a Bayesian posterior.
- Shrinkage MethodsShrinkage methods like ridge regression constrain coefficients toward zero to reduce variance, while dimension reduction techniques like PCR transform predictors into…
- Considerations in High DimensionsHigh-dimensional data, where features p exceed or equal observations n, causes classical least squares regression to fail by producing a perfect fit and severe…
- The Laplace MechanismThe Laplace mechanism ensures differential privacy by adding noise drawn from a Laplace distribution to a query's true output, with the noise scale calibrated to the…
- Advanced CompositionAdvanced composition provides a rigorous framework for analyzing cumulative privacy loss when multiple adaptive mechanisms are applied sequentially to the same dataset.
- Orthogonal PolynomialsOrthogonal polynomials form a basis for function spaces where distinct polynomials have zero inner product, enabling efficient least squares approximation without…
ccBYNC© 2026 SignalsLab. Lesson text shared as is under CC BY-NC 4.0.