The use of Bayesian Statistics in project work is well developed. Here's an example of the approach. The drivers for each uncertainty can be modeled to produce on estimate of the *risk* of the project not producing the desired outcomes.

What is needed is an understanding of the *prior* probabilities of the drivers of the probabilistic outcomes, the relationships between the events created by these networks.

**The Bayesian Network Approach ^{†}**

Bayesian networks provide decision support processes for a wide variety of problems where uncertainty and probabilistic reasoning is involved. The Bayesian Network is a *directed graph* with associated probability tables. The graph is the standard *nodes and arcs*. The nodes represent uncertain variables. Each node has a set of states that represent causal or influential relationships between the variables. There is a probability table for each node, providing the probabilities of each state of the variable.

For prior variables - variables without parents the table contains *marginal probabilities*. This is referred to as the *prior distribution* which represents the prior belief - the state of knowledge - for that variable. For each variable with predecessors (parents), the probability table has the conditional probabilities for each combination of the predecessor states. This is the *likelihood* function that represents how likely is a state of a variable given a particular state of its predecessor.

Bayesian Network are applied in situation that require statistical inference. The use in this instance knows some *evidence* - some variable states or events that have *actual *observations. These observed values represent the *posterior *probability of the event occurring. By applying Reverend Bayes rile in each affect node, they can influence other Bayesian Network node by propagating and modifying the probability distributions.

**Bayesian Networks Have A Natural Connection With Project Planning**

These networks can:

- Explicitly quantify uncertainty and model the casual relationships between the variables.
- Enable reasoning from effect to cause as well as from cause to effect.
- Make it possible to overturn previous beliefs in light of new data.
- Make predictions with incomplete data.
- Combine subjective and objective data.
- Arrive at decision based on visible and auditable reasoning.

**Bayesian Networks is a tool for decision support based on Estimating outcomes.**

So when we hear that decisions can be made in the absence of estimates, ask for tangible examples of how this can be done, the basis (mathematics) of these processes, and examples in specific domains of how this can be done. If no answer is forth coming, question the veracity of the claims.

**† "**Inference in Hybrid Bayesian Networks using dynamic discretisation," Martin Neil, Manesh Tailor, and David Marquez, Department of Computer Science, Queen Mary, University of London