Welcome to the Statistics for Data Science GitHub repository! This repository contains resources and materials for a course that aims to develop a strong understanding of statistical modeling, a fundamental skill for aspiring data scientists. Statistical models play a crucial role in answering research questions and extracting meaningful insights from diverse datasets.
In the Master of Information and Data Science (MIDS) program, grasping the content is just the beginning. This course is designed to help you not only understand statistical modeling but also cultivate essential skills for a successful data scientist:
By the end of the course, you will be able to fit and interpret statistical models, including linear and generalized linear models. This foundational knowledge will empower you to analyze data effectively and draw meaningful conclusions.
Learn how to map research questions and datasets to the appropriate statistical models. This skill is critical for identifying the right approach to tackle real-world problems and derive actionable insights.
Develop the ability to make careful and critical decisions about model building. You'll consider real-world implications and refine your problem-solving skills, a crucial trait of successful data scientists.
For this assignment, we dive into the Auto dataset, accessible here. Our approach follows a systematic path to gain insights and construct regression models. We initiated by exploring a data dictionary to understand variable meanings. We then built an initial regression model, focusing on 'mpg' as the dependent variable. Notably, we scrutinized the standard errors of the 'cylinders' predictor. Additionally, we performed data cleanup for 'cylinders' due to low observation counts for specific levels. Further analysis involved evaluating residual and QQ plots to assess model assumptions and identify potential violations. We also delved into model transformations, considering predictor variables such as 'displacement,' 'origin,' and 'cylinder levels,' along with the outcome variable 'mpg.' This structured approach guides us in extracting valuable insights and constructing robust regression models for the Auto dataset.
For this assignment, we analyze data from Dr. Kristen Gorman and the Palmer Station, Antarctica LTER, accessible here. Our approach involves systematic exploration, variable selection, and regression modeling. We start by exploring the dataset's structure using glimpse(). After identifying categorical and numeric variables and handling missing data, we select an outcome variable and a primary predictor. We create informative scatter plots and consider interaction terms. Regression modeling and interpretation follow, including assessing species significance. Finally, we evaluate adjusted R-squared values to draw meaningful insights from the data.