【正文】 t results obtained from the GDM are (1) the calibrated conditional probabilities。p(A n D) np(B n D)] (14b) = p(A n D) +p(B n D) 163。 and (3)the cost influence of starting attributes in a new project. As shown in Fig. 4, the calibrated results obtained from GDM are later used by the PWPCE model (stage 1) in module 2 of COMPASS to analyze a new project. MODULE 2 The general objective of module 2 is to analyze a new project by incorporating the past project performance data and the new project characteristics. The specific objectives of this analysis are to pute ? The active state probability of all of the attributes in a new project ? The probability of project cost escalation ? The cost influence of all of the attributes in a new project ? The total project cost escalation possible in a new project ? The probable project cost escalation The PWPCE model was developed to achieve these objectives. This model has four stages and it uses Bayesian analysis and the relative cost influence of attributes to pute the active state probabilities and the cost influence due to the active state of an attribute. The input data for module 2 are obtained from the DPM and the GDM of module 1 (refer to Fig. 5). The product of the active state probability and the associated cost influence of the attribute provides the probable cost influence of eac h attribute in a new project and also the probable escalation in the project cost. Probable Weighted Percentage Cost Escalation (PWPCE) Model The probability of an attribute attaining active state in a project is calculated in stage 1. This analysis is conducted by utilizing the Bayesian analysis techniques and the calibrated project performance data obtained from the GDM of module 1 (refer to part B of Fig. 5). Part B of Fig. 5 shows the analytical putations performed during stage 1 of the PWPCE model to determine the active state probability of attributes in a new project. The calibrated conditional probabilities and the calibrated active state probability of starting attributes A and B are obtained from the GDM. With this information, the joint and marginal probabilities of attributes are puted (Benjamin and Cornell 1970。 j =1 ... 10 (I1a,b) Group Decision Model (GDM) The objective of the GDM is to calibrate the information extracted in the DPM for use in analyzing a new project (refer to part B of Fig. 4). The desired calibration is achieved by soliciting the subjective (or judgmental) input from m project team members with respect to the new project characteristics. The subjective input provided by the team members is with respect to the following:The active state probability of starting attributes [Le., p(A = 1) and p(B = 1) in the example influence diagram] due to the influence of factors external to the system (refer to part A of Fig. 4) ? The conditional probability associated with an attribute and its preceding attribute(s), ., p(C = llA = 1), p(E = 11 C = 1), and so forth (refer to part B of Fig. 4) ? The cost influence of the starting attributes The bined information from the DPM and the GDM is used by the PWPCE model in module 2 to pute the probability of cost escalation and the percentage escalation in anew project of Data for Analyzing New Project As mentioned earlier, when calibrating data extracted from the past projects, it is important to take into account the unique characteristics of a new project. The process of calibration is carried out in four steps. In the first step, the primary decision maker (PDM) prepares a data collection sheet, as shown in part B of Fig. 4. The data analyzed in the DPM is included in the data collection sheet [marked I(PDM) in part B of Fig. 4, “conditional probability, perception of the team members].The data collection sheet is then forwarded to the other members of the team for their analysis and input. In the second step, the team members (., management, construction manager, design engineer, etc.) are required to analyze the information obtained from the PDM with respect to the new project characteristics。 p[(A =1) n (D =1)] (5) cost influence (CI) of attribute A on line itemj = CI(A)j (6) cost escalation (CE) in line itemj =(CE)j =CI(A)j + CI(DIA)j (7) CI(A187。 and (3) user input to establish threshold PWPCE value to isolate potential risk attributes by using the DAM and for developing a project cost control strategy. Several logical checks have been provided throughout the system to assist the user with data entry and analysis. MODULE 1 The objective of module 1 is to extract information regarding the conditional relationship attributes and their relative cost influence. This information is calibrated for use in a new project, with respect to the subjective input provided by the team members regarding the new project characteristics. Module 1 is prised of two models (refer to Fig. 2), the DPM and the GDM. Data Processing Model (DPM) The DPM has two stages (refer to Fig. 3). The objective of stage 1 of the DPM is to analyze the past project performance data provided by the user. This analysis establishes the conditional probability of an attribute attaining active state, given that the attributes preceding it in the influence pattern have attained the active state. The conditional probabilities calculated in this model are calibrated in the GDM. The calibrated conditional probabilities are later used by the PWPCE model in module 2 to pute the active state probability of attributes in a new project. The individual cost influence of attributes in each historical project is puted in stage 2 of the DPM (refer to Fig. 3).The process starts by isolating the necessary information from a number of past projects (say, n). Two criteria are remended for sel