This repository houses an ongoing series of small statistics projects. There will typically be three files for each R project; one will contain pure code (.R), another will be markdown (.Rmd), and the last (.pdf) can be viewed directly in github. The pdf file description includes the associated project description. There will be one file (.ipynb) for python projects as notebooks can be viewed directly in github.
This project explores and applies four methods of linear regression, in addition to a random forest model, to predict the sale price of homes based on 40 quantitative attributes (kaggle dataset link provided). The random forest model is included to evaluate the trade-off in time vs. potential increase in accuracy. The four linear regression methods applied within are Ordinary Least Squares (OLS), Ridge, Lasso, and Elastic-Net. Seeing as Ridge and Lasso are special cases of the Elastic-Net, an arbitrary alpha (.5) is used in the Elastic-Net model. This analysis also includes bootstrapped error bars of feature significance for each model. Data link is in the report.
Viewable output of one_factor.rmd file. This project explores (i) the assumptions for the one factor (one-way) ANOVA test using Citibike data, and (ii) various tools available in R to verify these assumptions prior to conducting the analysis. Some tools explored in this analysis are as follows: Shapiro-Wilk’s Normality Test, Levene’s Test for Equality of Variance, Tukey’s Honest Significant Difference Test, distribution visualization (histograms, boxplots), and further EDA. Citibike data was pulled and aggregated from here.