Example of pipreq on Jupyter Notebook:

pipreqsnb --savepath ./a3/python3.8.10_requirements.txt ./a3/a3.ipynb

Questions

• What is the difference between Shapiro-Wilk and Z-test? Both test for normality.
• A1. qqplot: How to add ci_mean (like in statsmodels) confidence interval?
• Why are Shapiro–Wilk W-statistic, t-statistic, F-statistic (when not 1) or Z-statistic shown, but we cannot use them directly? How p-values are calculated for these statistics?
• A3.1.4 fit4 and fit5: How to do orthogonal polynomials? Do I have the correct implementation? Which algorithm is used in R and how can it be reproduced in Python (with numpy, preferably)?
• A3. beat: How to use seaborn.pairplot() as matplotlib subplot (usually by supplying ax parameter)?
• A4.1.2.2: How to change the sign of the coefficients? Is it because of the order of values? Should it be fixed through contrasts' matrix encoding using C(Direction, contrast)?
• A4.1.2.4: How to use the R-like formula? There is no integration of sklearn with patsy.
• A4.1.2.8: How to set random seed in sklearn for KNeighborsClassifier? I can only do np.random.seed(1), but I do not think it is related.
• A6 Check the importance: How can I get IncNodePurity (increment in node purity) from sklearn.ensemble.RandomForestRegressor? I couldn’t find a way to get aggregated node purity increase.

Assignment Typos

• "Interprete" instead of "Interpret" in all places;
• A6 ggscatter(Carseats, x = "Sales", y = "Price" ...: axis labels were switched;
• A6.1.2.4 when building a random forest of classification trees: regression tees are meant because medv in Boston is of float type;

TODO:

• experiment to understand better A7.plot_svm;