Decision modeling

Minciardi R, Robba M, Sacile R (2007) Decision models for sustainable groundwater planning and control. Control Eng Pract 15 1013-1029... 

Trial validity is also grossly affected by the type of trial carried out. In schizophrenia there are several health-care decision models, retrospective mirror-image analyses (with or... 

Health-care decision models and uncontrolled mirror-image studies provide little in the way of meaningful results to inform clinicians and healthcare managers. This chapter focuses instead on the results of the more robust controlled mirror-image studies and prospective randomized, controlled trials. 

As already outlined, health-care decision models hold little water in the sophisticated environment of evidence-based medicine. Nevertheless, two UK evaluations (Davies and Drummond, 1993 Matheson et al, 1994) do give some insight into the outcomes of using clozapine in the UK National Health Service, although model data were largely derived from the USA. 

Spannheimer A, Clouth J, Gregor KJ (1999). Pharmacoeconomic evaluation of the rrearmenr of schizophrenia in Germany a comparison of olanzapine, risperidone and haloperidol using a clinical decision model. Poster presented at the ISPOR Second Annual European Meeting, Edinburgh, November 1999. 

Policy makers would benefit from a combination of strong field evidence of trends and well-established models to draw upon when assessing the benefits of past or future policy decisions. Models of mercury cycling and bioaccumulation are not yet adequately predictive across a range of conditions and landscapes. Results from a national mercury monitoring program, if carefully designed, offer the potential to... 

Col NF, Weber G, Stiggelbout A, Chuo J, D Agostino R, Corso P. Short-term menopausal hormone therapy for symptom relief an updated decision model. Arch Int Med... 

Yang J, Lee H (1997) An AHP decision model for facility location selection. Facilities 15 241-254... 

J. C. Ammons, M. J. Realff, and D. E. Newton, Decision Models for Reverse Production System Design, Handbook of Environmentally Conscious Manufacturing, Kluwer Academic Publishers, 2001, pp. 341-362. 

Let us take the incoming edges to node 3.1 of Chap. 3 as an example. We see that Sects. 2.3 on document models, Sect. 2.4 on work process models, and Sect. 2.5 on decision models are prerequisites for Sect. 3.1 on experienced-based design process support. 

In this contribution, we use the term decision model for both a representation of design rationale [856], i.e., the decisions taken by a designer during a design process that led to a particular artifact, and for a representation of some rules or methods which can guide a designer confronted with a decision problem or which can even solve a decision problem algorithmically. 

As indicated in Fig 2.1, the first four partial models are concerned with the representation of product data they can be considered as the origin of the product data models of the CRC, which are presented in the following subsections. The fifth partial model, in contrast, addresses work process modeling it forms the basis of the later-developed work process and decision models presented in Sects. 2.4 and 2.5. The partial models of VeDa s product data part are closely related to each other, and their mutual relations are explicitly represented in the model. The process part, on the other hand, is somewhat detached from the rest of the model as the interdependencies between the product part and the process part have not been explicitly covered by VeDa. 

In general, design decisions are too complex to be represented by a set of conditional branchings in a work process. Instead, a modeling approach tailored to decision modeling should by pursued (cf. Sect. 2.5). 

For the reasons discussed in Subsect. 2.4.5, the Process Ontology does not provide classes corresponding to Issue and Requirement within Process Models. Design decisions are now covered by a separate decision model (cf. Sect. 2.5). 

The InformationRelation is introduced as an abstract base class for any relation between Information items. Its usage is exemplified in the section on decision modeling. 

The necessity to incorporate goals makes notations like QOC or DRL a good starting point for a Decision Ontology. We have opted for the more expressive, but also more complex DRL. However, QOC can be regarded as a subset of DRL there is no necessity for a decision modeler to apply the entirety of DRL concepts when a simpler QOC-like representation is sufficient. 

The classes of the Decision module are shown in Fig. 2.29. All classes are derived from the abstract Decision Object. We first discuss the six classes located in the left part of the figure instances of these classes form the nodes of a decision model. Gray shapes inside the class boxes depict the graphical representation of their instances in the modeling examples below. 

Any statement in a decision model which may be subject to uncertainty or to disaccord, or, in general, may be evaluated, is a Claim. Claims are either SimpleClaims or relation classes derived from IsRelatedTo. Most of the relations introduced above are subclasses of IsRelatedTo, and thus they are actually Claims. Statements which cannot be represented as a relation between two DecisionObjects are modeled as SimpleClaims, typically qualified by a textual annotation. 

Integrating the work process and the decision model allows to remedy the difficulties in defining subclasses of Activity as they were encountered for the partial model Process Models of CLiP (called sub-activites there see Sect. 2.4.6). The relation class OutputInformationFlow defined in the Process Ontology does not impose any restrictions on the type of Information created in an Activity (see Fig. 2.31), whereas a DecisionActivity is required to produce at least one DecisionObject. Subclasses of DecisionActivity are characterized by... 

A decision model represented by means of the Decision Ontology contains much information that is also relevant when an MCDA method is used (e.g., the hierarchy of the Goals). The application of such methods would be simplified if this information could directly be used in an MCDA tool. However, MCDA methods impose restrictions on the relations between the Decision Objects. For instance, AHP requires the sub-goals of a goal to be independent of each other. These restrictions are formally defined within the DecislonAnalysis module. 

---