An Empirical Study on Statistical Analytical Procedures in Auditing

This empirical study has been done with the goal of developing auditing knowledge and the efficiency of its operations when using the statistical analytical procedures. In this research, eight alternative models have been evaluated, including five regression models, one time - series model (consus X-Ii) and two non-statistical models (Martingale and sub-Martingale). Both financial and non-financial data were collected from a sample of petrochemical companies for the period march, 1998 through March 2001. The information was used to predict sales revenue and production expense account balances. According to the results, regression models have better performance for predicting account balances in performing auditing analytical procedures in comparison to Two other models. Logarithmic regression has been evaluated as the best statistical analytical procedure. The foresaid procedure has a constant performance in sample companies of the industry. In performing statistical analytical procedures, monthly models perform better than seasonal ones. Pooled models have a better ability for prediction than single company models. Furthermore, the results of this research show incremental benefits of using nonfinancial variable in performing statistical analytical procedures in auditing.