Practical Predictive Analytics

Example – setting the outliers to NA

The next step is of course to start examining these outliers. For our example, the extremes are part of the random number generation process, therefore they are not really outliers. However, if you encountered this situation in your own data, you would begin to track down the reasons these extremes occur. Start by trying to associate these extremes with other data elements. Perhaps these outliers are appearing in certain age groups and not in others.

For our example, we will simply be setting the value to NA for these extreme values. We will also be creating a new variable, v1x, to house the new variable, and will not overwrite the value of the original variable. As you investigate new ways of detecting ...