Risk management - Uncertain events part 4
Uncertain events part 4
We have seen that in terms of uncertainty we can have:
Uncertainty in the value of an activity
This is the 3 point estimate [see Probability density function - Simplified version] of the value of a single activity which is represented in a model by the triangular PDF.
Uncertainty in an event
Normally, we are considering a single event. However, when an event can lead to more than one pathway we need to designate a likelihood for each [see Uncertain events part 2].
Correlated events.
We have seen [see Monte Carlo simulation output] that we can examine a number of activities over their 3 point ranges by employing random numbers to generate possible values for each activity. We would then aggregate all the individual values to get one possible total cost from many possibilities.
Using this technique we know that some activities will be near their MAXIMUM and others will be near their MINIMUM. The model reflects what we would expect in practice. This is good because a high in one area and a low in another, averaging over many activities, ought to reduce the RISK overall. There are swings and roundabouts. You gain on some activities and lose on others.
We have seen from the PDF curves [see Cost model part 2] that the likelihood of carrying out a project for the MINIMUM or MAXIMUM costs is very small compared to a value somewhere in between.
When looking at these activities we are assuming that they function independently and that there is no common theme between them.
However, many activities appear linked in that if one particular activity cost (or duration) is HIGH (or LOW) then a subsequent activity will also be HIGH (or LOW).
These events are known as CORRELATED events. See earlier examples [see full positive correlation], [see full negative correlation], [see independent correlation] and [see conditional correlation].
This might show itself in overestimating (or underestimating) the power needed for a heating and ventilation system. This could have a knock on effect in terms of pipe sizing, quantity, equipment and the site size for boilers etc.
The project manager needs to be aware of these correlated activities and make allowance for them in the model. If you don’t a HIGH might be cancelled out by a LOW in the model, where it should have been a HIGH as well, so generating a false total in the model.
Correlation makes extreme values more likely.