When building a cost estimation risk analysis model, how can we ensure validity when the data source is an expert’s opinion?

In a risk analysis model, the use of alternate parameters when your data ‘source’ is an expert’s opinion makes perfect sense. There is no need to justify the estimation of obscure parameters in the absence of data—simply estimate one or more percentiles instead. No more than three are needed. This is easily done using Alternate Parameters in @RISK’s define distribution window. I have found such a method especially appealing to clients in the cost estimation field when there is usually little to no data (relevant or otherwise) for parameter estimation.

However, by not explicitly defining the mean/standard deviation/minimum/maximum etc how do you know what they are? Are they still logical, or can the distribution now be sampled in an infeasible region? Essentially, are you aware of the implications of your choice of percentile parameters?

A tedious solution to this problem is to check each input distribution in your model one by one in the define distribution window and see for yourself what the min/mean/max etc are. For only one or two distributions this may seem reasonable. But how would you like to do that for a 1,000 line risk register? I’d straight out refuse! Instead, I’d add a couple of columns that calculate some useful theoretical statistics (the RiskTheo functions) I can use to sanity check my distributions.

An obvious example in cost estimation is the use of @RISK’s RiskTheoMin. When a variable cost is given a “Low” parameter (P10, say) rather than a strict “Minimum” (such as used by the ever popular Triangle and Pert distributions), the theoretical minimum is unknown. A column of RiskTheoMin functions with conditional formatting will quickly highlight which of your costs now have a negative theoretical minimum and need to be truncated etc. The same could also apply to maximums (RiskTheoMax). Further, checks such as “the probability the distribution will sample above/below X should be roughly Y%” can also be run with @RISK’s RiskTheoTarget functions.

Happy automatic model validity checking!

» Watch a short movie about Defining Distributions in @RISK

Rishi Prabhakar
Palisade Training Team

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s