Double-click on the 窶廸umber of chocolate bars窶 node to bring up the dialog box for this node. This time we will go to the 窶廛istribution窶 view, so we click the 窶廛istribution窶 icon on the left-hand side of the dialog box.
To reflect the uncertainty estimated in the marketing research, we will use a lognormal distribution where the 10%, 50% and 90% fractiles are approximately 7000, 10000 and 15000 respectively. Although it is possible to specify the distribution by inserting these fractiles directly, it is often more convenient to work with relative uncertainty. This way it is much easier to compare the uncertainties in different variables. We will however, illustrate both approaches: relative uncertainty and absolute uncertainty in this example.
In DynRisk this is done by separating the distribution into two factors: a scaling factor, referred to as the 窶徑ocal factor窶 and a relative uncertainty distribution. The actual value of the node, i.e., the value it gets during a Monte Carlo simulation, is obtained by multiplying the local factor with a value generated from the relative uncertainty distribution.