Journal of Official Statistics, Vol.8, No.2, 1992. pp. 167–182
Sensitivity Analysis of Empirical Studies
Abstract:Results from an empirical study depend on data, the statistical model, and the statistical techniques used. The statistical techniques are in turn constructed according to the model and a set of principles of statistical inference. Since statistical models are ultimately false, a source of uncertainty in the results is introduced. This uncertainty is distinguished from the sampling error, which traditionally is of primary concern in the statistics literature. Here, sensitivity analysis is defined as the investigation of how model misspecification and anomalous data points influence results. Included in sensitivity analysis are methods for assessing the influence of a particular source on the results, and suggestions about how to reduce unacceptably large influences. The purpose of this paper is to define some concepts of sensitivity analysis, illustrate the concepts in a few examples, and, perhaps most importantly, to emphasize the need for further research within this area.
Keywords:Assessing sensitivity; data penetration; model building; model formulation; principles of statistical inference; robustness; sensitivity function.
Copyright © Statistics Sweden 1996-2018. Open AccessISSN 0282-423XCreated and Maintained by OKS Group