Applied Biostatistics for Data Science Course

Lesson 1 • Variable Types and Measurement Scales

Distinguishes nominal, ordinal, interval, and ratio scales with biomedical examples. Correct classification drives every downstream analytical decision in the chapter.

Lesson 2 • Role of Statistics in Biomedicine

Defines biostatistics and its function in clinical and public health research. Establishes why rigorous quantitative reasoning is essential before any analysis begins.

Lesson 3 • Data Visualization Principles

Introduces histograms, box plots, and scatter plots as diagnostic and communication tools. Proper visualization reveals distributional assumptions before formal testing.

Lesson 4 • Data Quality and Preprocessing

Addresses missing data patterns, coding errors, and unit inconsistencies common in health datasets. Clean data is a prerequisite for valid inference in all subsequent chapters.

Lesson 1 • Bayes' Theorem in Diagnostics

Applies Bayes' theorem to update disease probability given test results. Connects prior prevalence to posterior probability, foundational for diagnostic test evaluation.

Lesson 2 • Core Probability Rules

Covers addition, multiplication, and complement rules with disease prevalence examples. These rules underpin sensitivity, specificity, and predictive value calculations ahead.

Lesson 3 • Continuous Probability Distributions

Covers normal, log-normal, exponential, and Weibull distributions relevant to biomedical measurements. Selecting the right distribution ensures valid parametric inference later.

Lesson 4 • Discrete Probability Distributions

Introduces binomial, Poisson, and negative binomial distributions for count-based health outcomes. Each distribution is matched to realistic epidemiological scenarios.

Lesson 1 • Multiple Testing and Error Control

Addresses family-wise error rate inflation and correction strategies for multi-endpoint studies. Proper correction prevents false discovery in high-dimensional biomedical data.

Lesson 2 • Logic of Hypothesis Testing

Defines null and alternative hypotheses, test statistics, and decision rules. Establishes the conceptual framework applied in every subsequent testing procedure.

Lesson 3 • Point Estimation and Confidence Intervals

Covers maximum likelihood and method-of-moments estimators alongside CI construction. Accurate interval estimation is the foundation of all inferential reporting.

Lesson 4 • Tests for Means and Proportions

Applies z-tests, t-tests, and proportion tests to common biomedical comparisons. Correct test selection depends on variable type and sample size covered earlier.

Lesson 1 • One-Way ANOVA and Post-Hoc Tests

Extends two-group comparison to multiple independent groups using ANOVA. Post-hoc procedures identify which specific group pairs differ after a significant F-test.

Lesson 2 • Nonparametric Tests for Continuous Data

Covers Mann-Whitney U, Kruskal-Wallis, and Wilcoxon signed-rank tests for non-normal outcomes. These rank-based methods are widely used in small clinical samples.

Lesson 3 • Assumptions Checking Before Testing

Covers normality tests, homogeneity of variance, and independence checks as prerequisites. Assumption violations guide the choice between parametric and nonparametric alternatives.

Lesson 4 • Two-Way and Repeated-Measures ANOVA

Introduces factorial designs and within-subject repeated measures for longitudinal biomedical data. Interaction terms reveal how treatment effects vary across subgroups.

Lesson 1 • Bias, Confounding, and Validity

Identifies selection bias, information bias, and confounding as threats to internal validity. Recognizing bias sources is essential for critically appraising published biomedical studies.

Lesson 2 • Randomized Controlled Trial Design

Covers randomization, allocation concealment, blinding, and intention-to-treat analysis in RCTs. RCTs provide the strongest evidence for causal inference when properly designed.

Lesson 3 • Observational Study Designs

Contrasts cross-sectional, cohort, and case-control designs by temporality and data structure. Design choice determines which effect measures are estimable and which biases arise.

Lesson 4 • Measures of Association and Impact

Computes risk ratio, odds ratio, rate ratio, and attributable risk for each study design. Impact measures quantify the public health burden attributable to a specific exposure.

Lesson 1 • Simple and Multiple Linear Regression

Derives OLS estimates and extends to multiple predictors for continuous outcomes like blood pressure. Regression coefficients quantify adjusted associations controlling for covariates.

Lesson 2 • Model Building and Variable Selection

Compares purposeful selection, stepwise, and penalized methods for covariate inclusion. Overfitting and underfitting are balanced through cross-validation and information criteria.

Lesson 3 • Confounding, Interaction, and Effect Modification

Distinguishes confounding from effect modification and shows how to model both in regression. Interaction terms reveal heterogeneous treatment effects across patient subgroups.

Lesson 4 • Logistic Regression for Binary Outcomes

Models the log-odds of a binary event such as disease presence using maximum likelihood. Odds ratios from logistic regression are the standard effect measure in epidemiology.

Lesson 1 • Log-Rank and Weighted Tests

Compares survival curves across groups using log-rank and Wilcoxon-Gehan weighted tests. Test choice depends on whether hazard differences are early, late, or proportional.

Lesson 2 • Parametric Survival Models

Fits exponential, Weibull, and log-normal accelerated failure time models to survival data. Parametric models enable extrapolation beyond observed follow-up for health economic models.

Lesson 3 • Cox Proportional Hazards Regression

Models covariate-adjusted hazard ratios using the semi-parametric Cox model. The proportional hazards assumption must be verified before interpreting any HR estimate.

Lesson 4 • Concepts of Survival and Censoring

Defines survival time, the hazard function, and right, left, and interval censoring mechanisms. Understanding censoring is essential before applying any survival estimator.

Lesson 1 • Generalized Linear and Mixed Models

Unifies regression for non-normal outcomes via link functions and exponential family distributions. GLMMs extend this to clustered count and binary outcomes in health data.

Lesson 2 • High-Dimensional and Omics Data Analysis

Addresses dimensionality reduction, penalized regression, and multiple testing in genomic and proteomic datasets. These methods handle the p >> n problem common in omics research.

Lesson 3 • Bayesian Inference for Biomedical Data

Introduces prior specification, likelihood, and posterior computation via MCMC for health models. Bayesian methods naturally incorporate prior clinical knowledge and quantify uncertainty fully.

Lesson 4 • Causal Inference Methods

Covers propensity score methods, instrumental variables, and difference-in-differences for causal estimation from observational health data. Causal framing goes beyond association.