Software Tools and Computational Issues

PRA models are usually implemented in a computer program, which can be a simple spreadsheet or more complex models in specific programming languages. This process may lead to errors. There simply is the possibility that the computer program does not perform as it was meant to. 

It must be ensured that the conceptual model is correctly translated into the mathematical notation, and that in turn into the computer code. Efforts to check for mistakes (test runs of example data sets, inspection of intermediate 

In modern software packages random number generators should work satisfactorily. However, if seeds are set manually, procedures should ensure that repetition of sequences is avoided. 

There are several sampling techniques in Monte Carlo analyses, the most common being random, median Latin hypercube and random Latin hypercube. Latin hypercube techniques are usually preferred because they need fewer iterations and thus are more efficient. They are, however, inferior to random sampling if high percentiles of the output are of interest and if the exact shape of the output distribution is important (Cullen and Frey 1999). 

The number of sampling iterations must be sufficient to give stable results for output distributions, especially for the tails. There are no simple rules, because the necessary number of runs depends on the number of variables entered as distributions, model complexity (mathematical structure), sampling technique (random or Latin hypercube), and the percentile of interest in the output distribution. There are formal methods to establish the number of iterations (Cullen and Frey 1999) however, the simulation iterations could simply be increased to a reasonable point of convergence. 

---