Applied Linear Regression Analysis (using R,SPSS,SAS,Python)

本课程教您如何根据您的需要，从最基本的级别到高级/专家级别进行线性回归分析。

我在本课程中采用的基本教学（或知识转移）哲学是，在学习如何应用这些方法通过软件进行数据分析之前，学生应该首先学习和理解“分析方法论的基本原理”。这与一些（类似）课程不同，这些课程的重点往往是教您如何使用软件运行回归分析（没有深入了解回归方法本身）。我的目的是让您首先掌握回归分析作为建模技术，并有信心解决任何需要线性回归建模的建模/预测问题。这意味着课程的第一部分在很大程度上独立于软件，

我相信一旦你学好这个重要的方法论部分，你应该能够使用任何软件进行应用回归分析。您会发现，线性回归分析的代码和步骤/过程在各种软件中非常相似（包括我们在本课程中使用的四个）。同样至关重要的是，回归分析的输出在大多数软件中在结构上非常相似且可以理解。因此，我的观点（以及采用这种方法的原因）是，如果您最初很好地理解了回归方法的基本原理，那么您应该能够随后使用任何软件并解释任何回归分析输出，并且还应该能够轻松地移动并使用各种软件（前提是您学习如何在该特定软件中使用或编码，

本课程的结构顺序如下。

1. 线性回归的介绍和动机。在这里，您将了解线性回归分析的概念和方法，并提供其应用的一些实际示例。我们还回顾了课程中使用的四个真实（公开可用）数据集，以演示如何在现实生活中应用线性回归分析。

2. 通过考虑简单线性回归 (SLR) 的情况，我们进一步了解线性回归的概念和方法。在这种情况下，您将构建模型以仅基于一个自变量或预测变量 (X) 来解释或预测因变量 (Y)。本介绍部分允许我介绍用于线性回归分析的术语和方法。我使用真实数据来演示方法论，在应用理论部分，我们使用 R 软件来演示方法论并解释软件的典型输出。对于那些感兴趣的人，我还（可选）介绍了用于简单线性回归 (SLR) 分析的普通最小二乘法背后的数学和统计理论基础。

3. 在学习了 SLR（简单线性回归）之后，我们转向更复杂的多元线性回归 (MLR) 情况，您现在可以根据几个自变量或预测变量（比如几个 X）对因变量 (Y) 进行建模或预测- 变量）。我们还学习了 MLR 的方法论（和一些理论），并将这些方法论应用于真实数据。我们再次使用 R 软件来演示回归分析和解释结果。对于那些对这方面感兴趣的学生，还有一些可选的讲座涵盖了用于多元线性回归 (MLR) 分析的普通最小二乘法背后的数学/统计理论。

4. 请注意，在 SLR 和 MLR 部分，我们涵盖的领域包括 (i) 良好线性回归分析所需的数据和建模要求，(ii) 回归分析所需的探索性数据分析，(iii) 估计参数/回归系数， (iv) 从软件输出解释回归分析的结果，(v) 使用可视化图和模型评估指标和度量进行建模后诊断，(vi) 关于回归系数的假设检验，并检查是否满足建模要求，(vii) ) 多元线性回归 (MLR) 分析的逐步回归分析。

5. 在介绍了本课程前面部分（小节）中的“回归方法”之后，我们将开始学习如何使用各种软件运行回归分析。我们在几个章节中进行了多次讲座，介绍和演示如何使用四种不同的软件进行线性回归分析。我们在单独的部分中介绍了此处使用的四种软件（R、SPSS、SAS、Python）中的每一种，其中（对于每个软件）我们演示了如何将其用于（i）绘制探索性图，（ii）运行回归建模，（ iii) 执行模型后诊断（视觉图、测量、度量、假设检验），以及 (iv) 在可能的情况下执行逐步回归分析。所有这些不同的软件都在几个真实的数据集上进行了演示，这些数据集展示了回归建模方法和结果的不同方面。需要强调的是，所有这些软件都具有不同的功能，这意味着我们所学的某些部分可能无法在特定软件中进行演示。供您参考，R 软件和 Python 都是免费的开源软件，通常通过直接代码在其中运行回归建模。另一方面，SPSS 和 SAS 是商业和专有（即非开源）软件。

From basics to advanced level training and practices in doing linear regression analysis using R, SPSS, SAS, Python

Genre: eLearning | MP4 | Video: h264, 1280×720 | Audio: AAC, 44.1 KHz
Language: English | Size: 6.52 GB | Duration: 13h 53m

What you’ll learn
Understanding how linear regression analysis works, including theoretical foundations, techniques, worked examples, live demonstrations of four software
Fundamentals and requirements for doing good linear regression, including data requirements, and tools for preliminary investigations (eg graphical plots)
How to use a variety of tools, measures, and metrics for evaluating if your linear regression model is a good fit for your data, and ways to improve the fit
Doing linear regression analysis in any of (or all) the four software covered, namely R, SPSS, SAS, Python. You will see learn from demonstrations of software
This course covers applied linear regression analysis fully. So you should not need to do a similar course again(except to learn to use a different software)

Description
This course teaches you how to do linear regression analysis from the very basic level, to advanced/expert level, depending on your needs.

A fundamental teaching (or knowledge transfer) philosophy which I have adopted for this course is that students should learn and understand the ‘fundamentals of the analytics methodology’ first, before learning how to apply those methodologies to do data analysis via software. This is different from some (similar) courses where the focus tends to be on teaching you how to use a software for running regression analysis (without deep understanding of regression methodology itself). My intention is for you to develop mastery of regression analysis as a modelling technique first, and have the confidence to tackle any modelling/prediction problem which requires linear regression modelling. This means that the first part of the course is largely software-independent, although I use R-software to demonstrate the concepts and also to help with your understanding and interpretation of software outputs for regression analysis.

I believe that once you learn this important methodological part well, you should be able to use any software for applied regression analysis. As you will find out, the codes and steps/processes for linear regression analysis are very similar across the various software (including the four which we use in this course). Critically important also, is that outputs from regression analysis are incredibly and understandably similar in structure, across most software. Hence, my view (and reason for adopting this approach) is that, if you understand the fundamentals of regression methodology well initially, you should then be able to subsequently use any software and interpret any regression analysis outputs, and also should easily be able to move across and use a variety of software (provided you learn how to use or code in that particular software, of course).

The course is structured sequentially as follows.

1. Introduction to, and motivation for, doing linear regression. Here you are introduced to the concept and methodology of linear regression analysis and some practical examples of its applications are provided. We also review four real (publicly available) data sets which are used in the course, for demonstrating how to apply linear regression analysis in real life.

2. We get further into the concept and methodology of linear regression by considering the simple linear regression (SLR) situation. This is the situation where you would be building a model to explain or predict a dependent variable (Y) based on just one independent or predictor variable (X). This introductory part allows me to introduce terminologies and methodologies used for linear regression analysis. I use real-life data to demonstrate the methodologies, and during the applied theory part we use R-software to demonstrate the methodologies and to explain the typical outputs from software. For those interested, I also (optionally) cover the basics of the mathematical and statistical theory behind the Ordinary Least Squares method for Simple Linear Regression (SLR) Analysis. Those not interested in that aspect (theory/mathematical/statistical methodology) can skip some of those optional Lectures.

3. After studying SLR (simple linear regression), we move to the more complex multiple linear regression (MLR) situation, where you would now be modelling or predicting a dependent variable (Y) based on several independent or predictor variables (say several X- variables). We also learn the methodologies (and some theory) for MLR, and apply the methodologies on real data. Once again we use R-software for the demonstrations of regression analysis and interpreting outcomes. There are also optional lectures which cover the mathematical/statistical theory behind the Ordinary Least Squares methodology for Multiple Linear Regression (MLR) analysis, for those students interested in that aspect.

4. Note that in both SLR and MLR Sections, we cover areas such as (i) data and modelling requirements necessary for good linear regression analysis, (ii) exploratory data analysis required for regression analysis, (iii) estimating parameters/regression coefficients, (iv)interpreting outcomes of regression analysis from software outputs, (v) post-modelling diagnostics using visual plots and model-assessment metrics and measures, (vi) hypothesis testing about regression coefficients, and to check that modelling requirements are met, (vii) stepwise regression analysis for multiple linear regression (MLR) analysis.

5. After covering the ‘regression methodologies in the earlier parts (Sections) of the course, we then move to learning how to run regression analysis using a variety of software. We spend many lectures in several Sections introducing and demonstrating how to use four different software for linear regression analysis. We cover each of the four software used here (R, SPSS, SAS, Python) in separate Sections, where (for each software) we demonstrate how to use it for (i) doing exploratory plots, (ii) running regression modelling, (iii) performing post-model diagnostics (visual plots, measures, metrics, hypothesis testing), and (iv) performing stepwise regression analysis, where possible. All these various software are demonstrated on several real data sets, which exhibit different aspects of regression modeling methodologies and outcomes. It is important to emphasise that all these software have different capabilities, which means that some parts of what we have learnt may not be able to be demonstrated in a particular software. For your information, R-software and Python are both free open-source software and one runs regression modelling in them typically via direct code. On the other hand SPSS and SAS are commercial and proprietary (i.e not open-source) software. Typically they have Graphical User Interface (GUI) which allows a user to run regression analysis purely by ‘point-and-click’ (if you cannot code in their proprietary code), but they can also run regression analysis purely via code.

This course will also allow you to check out and compare these various software, for regression analysis, if you are interested in that….and you don’t have to do several courses for that.

Who this course is for:
Statistical modellers, data analysts, data scientists, students, and researchers who want to properly understand how linear regression works in practice/applications, AND/OR people who are interested in learning how to do regression analysis using one or more of the software used in this course (i.e SAS, R, SPSS, Python).
People who are interested in understanding how the four different software (used in this course) are used for linear regression analysis