Chapter 9 Sensitivity analyses

9.1 Introduction

9.2 What happens if I remove outliers and influential observations?

Use the same examples you used in Section 6.10.5 etc. Material in the old Word document is still there in the MLR chapter

With and without the outlier(s), do your conclusions change? Depending on how much the estimates change, you can report this in various ways:

  • If nothing changes much, then you could simply note that you did a sensitivity analysis excluding the outlier(s) and the results changed only minimally.
  • If some things change a lot, but the conclusions don’t change, then you could mention more (e.g., if the slope decreases quite a bit, but remains significant).
  • If the conclusions DO change, then you should report both sets of results and note that it is not clear what the conclusion should be.

9.3 Bootstrap to check impact of assumption violations

Show how to use set.seed with bootstrap (right now, it doesn’t show exactly where to put the set.seed statement)

9.4 Writing it up