Sensitivity Analysis Resources


Less is more. Stay pure. Stay poor.
I was wondering if any one had a good reference for Sensitivity Analysis. Some thing fairly simple. I typically use SAS, but any general reference would work. I have seen people just look at the effects and do a quick calculation and say this much unknown misclassification or unknown reverse effects would have to exist to negate the results.

I never have enough time to review everything I want to, but just wanted to get a taste, if it is really as simple as I described.


Ambassador to the humans
What type of sensitivity analysis are you referring to? I typically see it in reference to testing how sensitive a bayesian model is to the selection of the prior distribution. I don't think that's what you're talking about here.


No cake for spunky
I know a variety of techniques in excel. These are what I [and many others] use for financial analysis. I don't have a link, but I can recommend books that discuss them. For instance we use them to show what the results will be if our assumptions are off by ten percent, twenty etc. Or to create scenarios. The methods are pretty basic.


No cake for spunky
Who knows ever what noetsi is talking about?? :(

I don't have my review, many pages long open right now - I will have to wait to Friday to pull it up. But two starting sources are "Excel 2010: Data Analysis and Business Modeling" by Wayne Winston 2011 pages 127-136 (this includes the use of tables for sensitivity analysis ) and 143-148 (this uses the scenario manager). A limit of this, of all excel approaches I know, is that you have to create the formulas that drive your model first. Solver and the various tables use these formulas to generate the result (they are essentially table/graphical features that utilize pre-existing formulas you create).

Scenarios are covered in 593-598 of the "Missing Manual" by Mathew McDonald 2010. Note that these are starting points. I had to flesh them out with lots of links. I will try to pull out that document which is much more complete on Friday.
One thing the author commented on was there was at least three ways to calculate R squared - which I had not known. I always thought it was SSR/SST.

Three take aways fron that link I think:

1) R squared is only explained variance with an intercept.
2) You calculate r squared differently with and without an intercept.
3) You really can not compare, at least in terms of R squared, a model with and without an intercept.

Interestingly the author notes, which I had not heard before that the assumptions that the conditional mean of the error term was zero was only a major issue without the intercept. I always assumed this was a critical issue with regression generally (well linear regression).

Pass your a+ course and comptia network+ exams in first try by using our northwestern passguide and and best quality Jacksonville University
Last edited: