Monday, December 4, 2017

Sensitivity Analysis

"It's what the bean counters call a simple actuarial analysis."
--Dr Getchell (Class Action)

Dilbert discusses sensitivity analysis. How sensitive is your model to changes in your assumptions?

When small changes in assumptions cause big or unpredictable changes to projected outcomes, then a model is overly sensitive. When a model weathers such changes well, it is 'robust.'

Consumers of forecasts, particularly economic forecasts, would be wise to apply sensitivity analysis to the underlying models that generate the forecasts.

Those who do will likely find that, more often than not, noise greatly exceeds signal.

No comments: