The most common response to this question is "Policy Summary!" The
DPL Policy Summary is a tool which makes it easier to understand what's going
on deep down in a complicated policy tree. Instead of showing you every branch,
it provides summary information about each node. For more information about
Policy Summaries, see the Decision Analysis Results chapter in the
DPL 6.0 Professional Getting Started Guide.
The second most common response and the subject of this article is the display
function. The display function is a general purpose tool in the Advanced Version;
you can use it to find out almost anything about a DPL analysis. As with most
general purpose tools, it isn't always obvious where to start.
How it works
The display function is most often used in get/pays, objective functions, and
node data. When DPL evaluates an expression containing a display function, it
prints something to the DPL Log. For example, if Unit_Cost is the name
of a value node, then
display(Unit_Cost, "%10.2f\n")
would cause DPL to print the value of Unit_Cost to the Log. The first
argument is the item to be printed; the second argument specifies the format.
"%10.2f" tells DPL to print a number with two digits to the
right of the decimal point and ten characters total. "\n"
indicates a carriage return, so the next number printed will appear on its own
line.
Display function format specifiers are the same as those used in the C programming
language. But don't panic -- you will almost always use one of the formats in
this article with only the numbers changed. If the format specifier is omitted,
DPL will use a default. The display function returns the value of its first
argument. This is handy when you are using it in a get/pay expression. Changing
a pay expression from Unit_Cost to display(Unit_Cost, "%10.2f\n")
doesn't change the model's results. When you don't want the value used in the
expected value calculations, you can multiply it by zero, as in 0*display(...).
Using the display function to selectively print information
Suppose you want to know the values of Unit_Cost in all scenarios
where you lose money. You can add a display function inside an @if
statement so that it is only evaluated when Profit is less than zero. In this
example, we put the display statement in a get/pay expression.
The output in the Log might look like:
3.25
3.78
Text in a format specifier that doesn't begin with % or \ is simply printed
out. In the above example, we could label our output by changing the format
specifier to "Unit_Cost = %10.2f\n". Then the Log would show:
Unit_Cost = 3.25
Unit_Cost = 3.78
Printing qualitative information
The display function can also be used to print character strings to the Log.
For example, you could sign your model by adding:
display("Bob's DPL model")
to a get/pay expression at the root of the tree (the node all the way to the
left).
The output in the Log would be:
08:12:16 Analyzing...
Bob's DPL model08:12:16 Complete
You can improve on this signature by adding a new line character to the string.
For example:
display("Bob's DPL model\n")
would give us
08:12:16 Analyzing...
Bob's DPL model
08:12:16 Complete
DPL's ability to print character strings to the Log is especially useful when
combined with the state-name function. Suppose you want to know the state of
Labor_Unrest in all scenarios where Unit_Cost is more than
five. You could replace the get/pay expression for Profit with:
Profit + @if(Unit_Cost>5, display(statename(Labor_Unrest),"%s\n"))
The format "%s\n" means print a string and then print a
carriage return. The display function returns zero when it's displaying a string,
so there is no need to multiply it by zero.
Display functions in the objective function
Even if you're not using multiple attributes, you may want to specify an objective
function which contains a display function. Objective functions can't contain
values which depend on the state of events, so you are limited, but the objective
function is the only place you have access to the path probability.
Suppose you want to know the probabilities of all money-losing scenarios. In
the objective function you refer to the first (or only) attribute as $1,
so if the only get/pay in the tree is get Profit, $1 is Profit and $1<0 means
we're losing money. The path probability is called $0 (in the objective
function only!). The objective function below prints the probabilities of all
money-losing scenarios:
