|
所在平台: Udemy |
课程主页: https://www.udemy.com/course/tabtrainer-minitab-multiple-regression-backward-elimination/
课程评论：没有评论
课程名称：Minitab中的多元回归 - Tabtrainer后向指导 课程概述：欢迎参加这门数据驱动的课程，属于Tabtrainer®认证系列——您可信赖的工业分析和应用回归建模平台。在本课程中，您将学习如何在Minitab中构建、优化和解释多元线性回归模型，使用来自Speedboard公司的实际生产案例。您将应用手动和自动后向消除的方法，识别最相关的预测变量，简化模型复杂性，同时保持统计完整性。 课程内容包括：相关性分析、基于方差膨胀因子的多重共线性检查、先进的模型诊断和最佳子集回归等。本培训将使您能够在工业质量、研发和过程优化方面做出自信的、基于证据的决策。该课程由Murad Mola教授主讲，他是TÜV认证专家、工业顾问，以及2023年德国“年度教授”，桥接学术深度与实际相关性。 在这门多元回归与后向消除课程中，参与者将学习： 1. 分析具有多个连续和分类预测变量的工业数据。 2. 应用后向消除法，解释p值、VIF与残差，并使用最佳子集回归简化模型。 3. 理解多元回归分析的基本原理，并将其应用于涉及连续和分类预测变量的实际工业数据。 4. 完整回归工作流，包括数据导入、探索、矩阵图，以及假设检验以评估初步趋势和关系。 5. 解释相关系数，并使用皮尔逊相关与p值确定变量间线性关系的统计显著性。 6. 评估各个预测变量（如甲板宽度、轮子硬度、甲板弯曲）对响应变量（最大速度）的影响。 7. 应用和解释方差膨胀因子（VIF）以检测和评估预测变量之间的多重共线性。 8. 逐步实施后向消除法，迭代地删除不显著的预测变量，简化模型并保持统计完整性。 9. 使用调整后的R平方和预测R平方评估和比较不同回归模型的拟合优度，确保模型的有效性和预测质量。 10. 通过残差分析评估模型假设，包括正态性、同方差性和独立性，使用“四合一”诊断图。 11. 执行自动后向消除法，并理解其相较于手动迭代消除的优势，尤其在高维模型中。 12. 应用最佳子集回归，识别在实践约束下最具影响力的预测变量，并解释先进的模型质量参数，如Mallows Cp、PRESS、AICc和BIC。 此课程致力于将理论与实践相结合，为参与者提供坚实的分析基础和实用技能。
Welcome to this data-driven course from the Tabtrainer® Certified Series - your trusted platform for industrial analytics and applied regression modeling.In this course, you'll learn to build, refine, and interpret multiple linear regression models in Minitab, using a real production case from the Speedboard Company. You'll apply both manual and automated backward elimination to identify the most relevant predictors, reduce model complexity, and maintain statistical integrity.From correlation analysis and VIF-based multicollinearity checks to advanced model diagnostics and best subsets regression, this training equips you to make confident, evidence-based decisions in industrial quality, R & D, and process optimization.Led by Prof. Dr. Murat Mola, TÜV-certified expert, industrial consultant, and Professor of the Year 2023 in Germany, this course bridges academic depth with practical relevance under the trusted brand Tabtrainer®.The course Multiple Regression with Backward Elimination teaches participants how to:Analyze industrial data with multiple continuous and categorical predictors. Apply backward elimination, interpret p-values, VIFs, and residuals, and use best subsets regression for model simplification. Emphasis is placed on practical model optimization and real-world decision-making:Understand the basics of multiple regression analysis and apply it to real-world industrial data involving both continuous and categorical predictors.Conduct a full regression workflow including data import, exploration, matrix plots, and hypothesis testing to assess initial trends and relationships.Interpret correlation coefficients and determine whether linear relationships between variables are statistically significant using Pearson correlation and p-values.Evaluate the effect of individual predictors (e.g., deck width, wheel hardness, deck flex) on the response variable (maximum speed) using p-values and model coefficients.Apply and interpret the Variance Inflation Factor (VIF) to detect and assess multicollinearity between predictor variables.Perform step-by-step backward elimination, removing non-significant predictors iteratively to simplify the model while preserving statistical integrity.Use adjusted R-squared and predicted R-squared to evaluate and compare the goodness-of-fit of different regression models, ensuring model validity and predictive quality.Assess model assumptions through residual analysis, including normality, homoscedasticity, and independence, using "Four-in-One" diagnostic plots.Execute automated backward elimination and understand its benefits compared to manual iterative elimination, especially in high-dimensional models.Apply best subsets regression to identify the most influential predictors under practical constraints and interpret advanced model quality parameters such as Mallows Cp, PRESS, AICc, and BIC.
