Use probability distribution methods. When you understand the likelihood of a risk event happening, you will understand the potency of such events in causing cost to vary and by how much.

I should evaluate the correlation between index priority risk (Severity index X Frequency Index X Detectability Index = Estimating severity index) and direct and indirect costs

These references are from my company's web site. I think they will help. 

Book : http://broadleaf.com.au/resource-material/project-risk-management-guidelines/

Videos: http://broadleaf.com.au/resource-material/video-resources/

Papers: http://broadleaf.com.au/resource-material/project-cost-contingency/ , http://broadleaf.com.au/resource-material/weaknesses-of-common-project-cost-risk-modelling-methods/

Webinar: https://pmi.adobeconnect.com/_a821822776/p34es5w4wcg/?launcher=false&fcsContent=true&pbMode=normal

Let me know if you have specific questions

Great resources!!! Please accept my many thanks for your invaluable responses 

No problem - you might find other materials on the site useful and, as I said earier, let me know if you have further questions

I am sorry if this is a silly question but Stephen, how effective do you think the risk assessment model can help to justify a particular construction solution to a project by assessing the risk profile behaviours with and without the solution or between various options? (e.g. to decide to adopt prefabrication or not) 

I often do this, not usually for construction projaect strategy but for IT bid selection. The principle is the same. See http://broadleaf.com.au/resource-material/tender-price-risk/

We carry out the risk assessment on the options, in your case the project assuming one strategy and the project assuming the other strategy, then compare their cost distributions. It's not always a simple comparison as you might find one has a low mean with a wide distribution that extends into some very high values that represent a major exposure while the other has a higher mean with a smaller spread so even its worst case wouldn't be as bad as the other one but the broader distribution might offer the possibility of a lower outcome if things go well.

People often simplify this by using their chosen risk level for the contingency, perhaps the P80 or P90, as the basis for the comparison. Sometimes though they prefer the predictability of the narrow distribution as a sign that this option will be easier to control. It all has to be passed back into the common sense filter at the end

I understand that the higher the construction risk  will mean a higher probability of that risk affecting the total construction cost. It also means severity and frequency are high. I would advise to centre your investigation on impact severity and frequency/ likelihood. You can also use mathematical probability modelling to to determine likelihood as stated by first contributor to your question

Further to the above replies, I would add that you can use Monte Carlo to quantify the risk estimations.  That will provide you with a probability distribution showing best case, worst case, mean and most likely aggregate cost range.  As you refine your assumptions, the distribution can be recalculated.  I attach material I wrote on this topic. 

Article Playing risk roulette with Monte Carlo simulation using Micr...