2

I have seen and heard people use "sensitivity analysis" to refer to both:

  1. How different values (e.g., just males or the whole cohort) of an independent variable affects the model, and
  2. How the addition or removal of an independent variable in a multivariable model affects model fit.

In addition to the answer on: https://stats.stackexchange.com/a/194291/112640

Does sensitivity analysis refer only to altering assumptions or also variables?

I wonder if anyone could clarify what people actually mean when they refer to 1) and 2) and why they're confused with sensitivity analysis, and if anyone could give a worked example of sensitivity analysis, for example using Stata or alternatively provide some intuitive references.

Community
  • 1
bobmcpop
  • 1,063
  • 1
  • 14
  • 20
  • 1
    it's basically a partial derivative with respect to a parameter or input in a very broad sense. for instance, it could measure the impact of the coefficient change 10% up or down. it's a way to measure the model risk. if your forecast or output is very sensitive to a coefficient it tells you that there's a high model risk. – Aksakal Dec 20 '16 at 18:41
  • 1
    See introductions of http://www.andreasaltelli.eu/file/repository/intro_v2fin.pdf and https://hal-univ-tlse3.archives-ouvertes.fr/hal-00975701/document – Pop Dec 22 '16 at 13:14
  • @Pop https://link.springer.com/content/pdf/10.1007/978-3-319-12385-1_31.pdf this might be a better formatted version of your first link – Doc Brown Jan 14 '19 at 11:15

0 Answers0