07. Groups and Outliers

How Groups can effect 'r'

Groups can cause a relationship to be stronger, weaker, or have little effect. Investigate the influence of groups.

If there are groupings in your data that behave differently then consider fitting a different linear regression line for each grouping to test the influence of the group.

How Outliers (Unusual Values) can effect 'r'

The relative position of an Unusual Value can have a huge effect on the 'relationship'.

Investigate the affect of Unusual Value and the robustness of the model. Simulation, move the unusual value, (allow macros)

They may increase the strength of any relationship

They may reduce the strength of any relationship

They may have no influence on the strength of any relationship

Remember the CONTEXT!!!!

A good approach to modeling bivarate data is to complete the analysis including unusual values, and also analyse the data without unusual values. A comparison of the two models can be done, and a discussion of the significance, nature and appropriateness of the unusual values can be made.