Modeling, Analysis and Display Tools
Model Building
TreeAge Pro employs a structured methodology that breaks the problem
into its component
parts, distinguishing between the available decision options and the
uncertainties associated with each. Based on this analysis, a model
in the form of either a decision tree or an influence diagram is graphically
displayed on screen, and the user interacts with it.
Many people prefer to structure a problem as a decision tree. Others
find that building an influence diagram helps them to think through
the problem. With TreeAge Pro you retain the flexibility to decide
whether a particular problem is better modeled as a tree or an influence
diagram. If you start with an influence diagram, TreeAge Pro can automatically
convert it into the equivalent, asymmetrical decision tree - and having
both forms of graphical output available can facilitate the often
critically important task of communicating both your view of the problem
and its proposed solution.
TreeAge Pro imposes no artificial restraints on the size or content
of models. Unlike other software packages, there are no built-in limitations
on the number or location of variables, nodes, decision options, chance
events, etc. Of course, in light of the graphics involved, working
with very large models can be expected to require substantial resources,
both RAM and CPU.
Analysis
TreeAge Pro calculates the expected value of each scenario, identifies
the decision maker's optimal strategy, and enables testing and analysis
of the recommendation. Sensitivity analysis makes it possible to identify
those uncertainties which - in light of the potential range of outcomes
- are not critical to the decision at hand; for each uncertainty where
further analysis and expense might be warranted, TreeAge Pro's value-of-information
calculations will specify the most which should be spent in an effort
to reduce or eliminate the uncertainty.
Analytical features include 1, 2 and 3-way sensitivity analyses,
tornado diagrams, threshold analysis, single and comparative probability
distribution graphs, Bayes' revision, Monte Carlo simulation and multi-attribute
analysis. Markov processes and cost-effectiveness analysis are available
through the TreeAge Pro Healthcare module.
Communicating Results
In addition to being able to display models graphically as
both influence diagrams and decision trees, TreeAge Pro offers a wealth
of supporting text and graphical reports.
Sensitivity analysis is used to test the sensitivity of a proposed
course of action to changes in the
projected value of one or more variables. For each variable, a range
of values is substituted for the point value used in the model.
If only a single variable is being tested - known as a 1-way sensitivity
analysis - the results of the analysis will be represented in a line
graph. When the analysis is performed at a decision node, the expected
value of each decision option will be plotted as a function of the
range of values assigned to the selected variable.
Any decision option represented by a horizontal line is unaffected
by the selected variable. Deviations of the line from the horizontal
indicate changes in the expected value of that alternative due to
changes in the value of the selected variable.
If two or more decision alternatives have the same expected value
at a given value of the selected
variable, this will be represented on the graph by intersecting lines.
Crossing points in the graph that also represent a change in the optimal
strategy are called thresholds. TreeAge Pro will specify the intersecting
values at each threshold.
In a 1-way sensitivity analysis, all other values in the model are
customarily held constant while the value of the selected variable
is varied across a specified range. However, in some cases a factor
that causes uncertainty in the value of the selected variable will
also affect one or more other values, making it unrealistic to leave
these values unchanged during sensitivity analysis.
TreeAge Pro makes it possible to specify direct or inverse correlations
among values. If any one of the correlated variables is selected for
sensitivity analysis, the values of its correlated partners will also
be varied, and this will be specified by a graph legend.
TreeAge Pro offers a second, more precise form of threshold analysis.
The result of this analysis is a detailed, text report describing
how the optimal strategy is affected by changing the value of a single
variable across a designated range.
It's also possible to perform sensitivity analysis on multiple variables.
A 2-way sensitivity analysis tests the sensitivity of a proposed decision
to simultaneous changes in the values of two, independent variables.
The values of any variables correlated with either of the selected
variables will also be varied, in accordance with the specified terms
of the correlation.
Results of the analysis are presented in a region graph in which
regions of different colors are used to indicate the optimal choice
at any given values of the two variables.
A 3-way analysis involves three variables and three ranges of values.
Results are depicted in an
animated region graph, which resembles the 2-way region graph but
adds a slider bar to change the value of the third variable. With
each change in the value of the third variable, the graph displays
a snapshot of the 3-way analysis specifying how the 2-way analysis
of the first two variables is affected by varying the value of the
third variable.
Another type of graphical sensitivity analysis is called a tornado
diagram. It is a set of 1-way sensitivity analyses combined in
a single graph. It can include all or a subset of the variables defined
in a model. A separate range is assigned to each. A set of correlated
variables is depicted as a single, composite variable.
In the resulting graph, a horizontal bar is generated for each variable
being analyzed. Expected value is displayed on the horizontal axis;
each bar represents the range of expected values generated by changing
the value of the related variable. Thus, a wide bar indicates that
the associated variable has a potentially large effect on the model's
expected value. The bars are arranged with the widest at the top and
narrowest at the bottom, giving the graph a funnel-like appearance.
Reliance on expected values calculations serves to maximize the
likelihood of a good outcome. However, it can neither guarantee a
favorable outcome nor rule out a highly unfavorable one. To guard
against the latter, TreeAge Pro offers several methods for analyzing
the risks inherent in the decision options. In many cases risk analysis
will confirm the option with the best expected value; in others it
may suggest choosing a potentially less valuable, but also less risky,
option.
To assess risk TreeAge Pro will, for each decision option, specify
the standard deviation, the range of potential outcomes, and the probability
of an outcome above or below any target value. In addition, TreeAge
Pro will graph the probability distribution of a single decision option
or create a graph which compares the probability distributions of
multiple options, making it possible to identify any instances of
deterministic or stochastic dominance.
For information on the additional reports - both graphical and text-based
- included in the
TreeAge Pro Healthcare
Module, click here. >
For information on the additional reports - both graphical and text-based
- included in the
TreeAge Pro Excel
Module, click here. >