The results of any epidemiological study are limited both by the data that are available and, on the face of it, the existence of a mathematical approach to solving the problem that the data are desired to address. In this final chapter of the book we shall consider three approaches to generating ‘quasi-data’ that use the observed data to construct data that could have been found under similar conditions. This approach offers relief from the mathematical constraints of classical statistics and allows for an extended range of inferences. In each case this is achieved through exploitation of the immense computational power and speed of modern computers. These three approaches are:

• The bootstrap, which is ideally suited for problems of estimation – precisely, expressions of confidence intervals (CIs). This is covered in Section 14.2 to Section 14.7.