Performance measures

To compare the performance of the statistical method in the different simulation conditions, several measures can be used. The suitability of the methods depend on the simulation design and the method that is to be validated. 


The most important measure is bias. Bias is the systematic deviation of an estimate from the reference. The bias can be calculated as the absolute deviation or as a percentage. The latter form is called standardized bias. 


When bias is calculated over multiple simulation iteration and the average is taken, the negative values can cancel out the postive values. This is not always a problem, this may just show that there is no systematic error. However, if you want to have an idea of the accuracy of the method as well it may be useful to look at the squared bias (i.e. squared error). The average over all squared errors (i.e. mean squared errors) provides information of the accuracy and efficiancy of an estimate.