How to Automatically Remove Outliers in Regression Calculations
When running Regression outputs or Generalized Linear Models, you may get a warning stating "Unusual observations detected." If that's the case, then it is possible that you have outliers in your data which are having an adverse effect on your model.
Process
To automatically remove these outliers, take these steps:
1. Create and run your Regression output as usual (see [Create Regression] for more on the different types available).
2. If there are unusual observations in your data which may be caused by outliers, you'll get a warning, like this:
3. Select your regression output in the Pages pane in Displayr (or in the report tree in Q), and copy/paste it. Work with the copy from here on.
4. Select the copy, and go to the settings for the regression analysis. Under Inputs > Automated outlier removal percentage specify a percentage limit for the removal of outliers.
5. Re-run the regression to see if removing the outliers made a difference to your outputs. If the model improved, then removing the outliers has been successful.
If the model still shows the warning after re-running it and removing outliers, then consider increasing the value. However, do note that the more outliers you remove, the less sample you'll have in your model which may result in other consequences.