|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/fundamentals-of-bayesian-statistics/
课程评论：没有评论
课程名称：应用贝叶斯数据分析与R 课程概述：本课程将带您深入贝叶斯统计的世界，探索这一在不确定性中推理的强大而直观的框架。通过使用真实数据和在R中的动手编码，您将学习如何使用先进工具（如brms和Stan）构建、评估和解释贝叶斯模型。不论您是希望深化统计工具的数据显示科学家、希望建模复杂效应的社会科学家，还是对贝叶斯推理抱有好奇的初学者，本课程旨在为您的分析带来清晰、信心和能力。 课程开始时，我们将重新塑造您对概率的思维方式。您将学习将概率视为信念的程度，而不是长时间的频率，这种观点自然而然地引导您进入贝叶斯推理的世界。通过直观的解释和基于代码的演示，我们将探讨如何利用新数据更新先验信念，从而形成后验结论。 接下来，我们将进入现代贝叶斯分析的核心计算方法。您将掌握马尔科夫链蒙特卡洛（MCMC）抽样，从Metropolis-Hastings算法开始，逐步深入到哈密顿蒙特卡洛（HMC）及NUTS等算法，这些算法驱动着现代贝叶斯引擎如Stan。 但是，仅有理论是不够的。本课程将提供丰富的实际应用，使用强大且易于使用的brms包，这是一个Stan的前端，允许您使用熟悉的R语法拟合复杂模型。您将构建贝叶斯线性回归、模拟数据、检查假设，并像专家一样解释您的结果。 对于准备深入学习的学员，我们将深入Stan模型的直接编写，使您完全掌控模型结构、似然和先验。您将探索从简单线性模型到非线性增长曲线及层次结构的所有内容。 同时，我们还将为您提供模型验证和选择所需的工具，包括后验预测检查、交叉验证和期望对数预测密度（ELPD）。您将学习如何诊断收敛问题、识别不同的过渡，并从模型制定到决策制定遵循一个原则性的贝叶斯工作流程。 课程结束时，您将能够： - 像贝叶斯人一样思考，结合先验知识并利用数据更新信念。 - 自信地使用MCMC方法拟合和诊断贝叶斯模型。 - 进行贝叶斯回归分析，了解何时以及如何使用Stan自定义模型。 - 理解如何从先验和后验中进行模拟，检查模型拟合并清晰地传达不确定性。 - 将一个原则性的贝叶斯工作流程应用于现实数据问题，从数据探索到最终模型验证。
This comprehensive course will take you on a journey into the world of Bayesian statistics, one of the most powerful and intuitive frameworks for reasoning under uncertainty. Using real data and hands-on coding in R, you'll learn how to build, evaluate, and interpret Bayesian models using cutting-edge tools like brms and Stan.Whether you're a data scientist aiming to deepen your statistical toolkit, a social scientist wanting to model complex effects, or a beginner curious about Bayesian reasoning, this course is designed to bring clarity, confidence, and capability to your analysis.We begin by reshaping how you think about probability. Rather than treating it as a long-run frequency, you'll learn to think of probability as a degree of belief-a perspective that naturally leads to Bayesian reasoning. Through intuitive explanations and code-based demonstrations, we'll explore how prior beliefs can be updated using new data to form posterior conclusions.From there, we move into core computational methods that allow modern Bayesian analysis to scale. You'll master Markov Chain Monte Carlo (MCMC) sampling-starting with the Metropolis-Hastings algorithm and moving toward Hamiltonian Monte Carlo (HMC) and NUTS, the algorithms that power modern Bayesian engines like Stan.But theory alone isn't enough.That's why this course is packed with practical, real-world applications using the powerful and user-friendly brms package in R-a front-end to Stan that lets you fit sophisticated models using familiar R syntax. You'll build Bayesian linear regressions, simulate data, check assumptions, and interpret your results like a pro.For those ready to go deeper, we'll open the hood and dive into writing models directly in Stan, giving you complete control over model structure, likelihoods, and priors. You'll explore everything from simple linear models to non-linear growth curves and hierarchical structures.We'll also equip you with the tools needed for model validation and selection, including posterior predictive checks, cross-validation, and Expected Log Predictive Density (ELPD). You'll learn how to diagnose convergence issues, identify divergent transitions, and follow a principled Bayesian workflow from model formulation to decision-making.By the End of This Course, You Will:Be able to think like a Bayesian, incorporating prior knowledge and updating beliefs using dataConfidently use MCMC methods to fit and diagnose Bayesian modelsPerform Bayesian regression analysis using brms, and know when and how to customize models using StanUnderstand how to simulate from priors and posteriors, check model fit, and communicate uncertainty clearlyApply a principled Bayesian workflow to real-world data problems, from data exploration to final model validation
