Bayesian Belief Networks (BBN) Experimental Protocol
This chapter provides an introduction to BBN experimental protocol, the experimental protocol for BBN, the characteristics of a random experiment, and the conduct a Bayesian experiment, which includes the following 11-Steps: (a) Step 1: identify a populat
- PDF / 132,100 Bytes
- 6 Pages / 439.37 x 666.142 pts Page_size
- 63 Downloads / 208 Views
Bayesian Belief Networks (BBN) Experimental Protocol
4.1
Introduction
This chapter represents the statistical methodology I followed in formatting the example Chaps. 5, 6, 7, 8, 9, 10, 11, 12, 13 and 14 that follow. My intent is to present a starting point reference guide for naı¨ve researchers when exploring and learning BBN.
4.2
BBN Experimental Protocol
This chapter provides a road map to the required statistical and probability theory review to cover the multi-dimensional and hierarchical relationships that exist in BBN.
4.3
Characteristics of a Random Experiment
The follow-on chapters contain ten random example experiments across multiple areas of research interest. Using the statistical definitions defined in Chap. 3, Statistical Properties of Bayes’ Theorem, each random experiment contains two or more events and elements where I derived the data from random sampling techniques using Monte Carlo simulations. These examples are fictitious in nature but do represent reality at a higher levels of thought and the techniques and procedures I outline here can be easily adapted to similar experiments. To do this, it is critical that researchers understand the premise of conducting the random experiment so they can be grounded in theory. This chapter suggests a methodology as a starting point for the follow-on chapters and in conducting subsequent independent research. J. Grover, Strategic Economic Decision-Making: Using Bayesian Belief Networks to Solve Complex Problems, SpringerBriefs in Statistics 9, DOI 10.1007/978-1-4614-6040-4_4, # Springer Science+Business Media New York 2013
43
44
4.4
4 Bayesian Belief Networks (BBN) Experimental Protocol
Bayes’ Research Methodology
There exists a universe of events that have actually occurred in the past that are invisible to us. We have not seen them but we know they exist. Having a scientific mindset, we desire to determine an acceptable truth of the proportions that the elements of these respective events represent. We begin this process by making an initial assumption of these proportions based on our beliefs. We then conduct an experiment by making random draws from a population of interest to determine these proportions. In the language of Bayes’ statistics, we refer to this population base as the universe of all possible events in which we seek to make their respective elements visible or known. In the Bayesian universe, we seek to determine conditional relationships across these events. In essence the unobserved event then becomes our “Cause” and the observable events our “Effect” and vice versa. In general, in doing this we seek to identify a universal set. Within this universe, we also seek mutually exclusive (disjoint) sub-sets. The unobservable sub-set contains the elements we desire to discover. We discover these elements by conditioning the observable on the unobservable event(s) and count these frequencies of dependencies. In essence, we are asking what is the probability of an Event B given the evidence of an Event A, or PðBjAÞ in a BBN. Under
Data Loading...
