A frequent error in risk analysis is that interdependences between variables are often ignored, which may result in the under-estimation of total risk. For example, the project team may include the cost of insuring a risk in the PSC, but if the event that a claim were to be made, the insurance premium to reinsure that risk in following periods may increase. Thus the occurrence of an identified risk may trigger a further risk (increased insurance costs) that are directly correlated with occurrence of the initial event.
Correlation analysis is a complex and time-consuming process. The analysis should only be undertaken where the cost is material and would impact significantly on the total cost of the PSC. However, correlation analysis must be undertaken where it is readily apparent that interdependences exist between key variables.
Initial tests of correlation can be constructed by using data analysis functions such as the CORREL and PEARSON add-in functions in EXCEL. The CORREL function is used to define rank and order coefficients for non-linear relationships, while PEARSON calculates the Pearson's coefficient for the degree of linearity between two variables. The correlation coefficients are defined between +1, (the variables move together), and -1, (the variables move in opposite directions). Ideally, the variables in question should have a correlation coefficient close to 0, indicating perfect independence.
Sophisticated software packages are often required to undertake a comprehensive correlation analysis and are usually included in Monte Carlo simulation software, discussed in the following section.