Changing Probabilities (Command Program)

The Command Program allows you to examine the effects of different probability distributions for the chance events in your problem. You can use the let command in the same manner as in Changing Values. Using the let command, you can modify the probabilities assigned to chance events, which allows you to study the effect on the outputs of your model.

There are two important guidelines to follow when changing the probabilities of chance events in the Command Program.

- The probability distribution for each state of the conditioning event(s) must be specified separately.

- The entire distribution must be defined under each specification.

These guidelines will become clear in the following examples. Continuing with the probability example defined in Displaying Probability Distributions, the following let command modifies one probability distribution for chance event MarketShare. (Because event MarketShare has two states, only one probability value is needed to specify the distribution.)
let MarketShare | Competition.Strong = {.45};

The Session Log displays the modified marginal probability distribution of the chance event MarketShare.
-MarketShare.Gain: 0.5
-MarketShare.Lose: 0.5

Each probability specification must be changed by a separate let command. For instance, if you also want to modify chance event MarketShare where Competition is in state Weak, you must write another let command.

Consider the following chance events in another example:

chance NightShift = {.45};
chance CycleTime.{4} | NightShift={.2,.3,.4},{.15,.26,.5};

Suppose you wanted to modify some of the probabilities for event CycleTime, specifically from .4 to .33 in the first set of brackets and from .15 to .1 in the second set of brackets. Remember you must specify the distribution corresponding to each conditioning event state separately and define the distribution in its entirety. You can make these modifications with the following commands.

let CycleTime | NightShift.s1 = {.2,.3,.33};
let CycleTime | NightShift.s2 = {.1,.26,.5};

To see what impact the changed probabilities have on the optimal Policy Tree, execute the command, and then choose Run | Decision Analysis. The Policy Tree and outcome distributions this run generates will reflect the modified probabilities.

These changes are now part of the compiled code in memory; to remove the modifications, you must recompile the model.

Versions: DPL Professional, DPL Enterprise, DPL Portfolio

See Also

Running a Decision Analysis

Working in the Command Program