Capping the variable at arbitrary maximum and minimum values

Similarly to winsorization, we can replace the extreme values by values closer to other values in the variable, by determining the maximum and minimum boundaries with the mean plus or minus the standard deviation, or the inter-quartile range proximity rule. This procedure is also called bottom and top coding, censoring, or capping. We can cap both extremes of the distribution or just one of the tails, depending on where we find the outliers in the variable. In this recipe, we will replace extreme values by the mean and standard deviation or the inter-quartile range proximity rule, using pandas, NumPy, and Feature-engine, and using the Boston House Prices dataset from scikit-learn.

How to do it...

...