Simulation Input Modeling in the Absence of Data

Input Model;Simulation

Liebsch, Cindy Marie ; Dr. Reid Ness, Committee Member,Dr. Jerome Lavelle, Committee Member,Dr. Stephen Roberts, Committee Co-Chair,Dr. David Kaber, Committee Co-Chair,Liebsch, Cindy Marie ; Dr. Reid Ness ; Committee Member ; Dr. Jerome Lavelle ; Committee Member ; Dr. Stephen Roberts ; Committee Co-Chair ; Dr. David Kaber ; Committee Co-Chair

University:North Carolina State University

关键词: Input Model;    Simulation;   

Others  :  https://repository.lib.ncsu.edu/bitstream/handle/1840.16/278/etd.pdf?sequence=1&isAllowed=y

美国|英语

来源: null

【 摘 要 】

Simulation models provide a powerful tool for analyzing real-world systems. These models are driven by input data, so when inputs are unknown and no data exists, the development of the simulation model becomes problematic. This research addresses the problem of modeling inputs in the absence of data, with the goal being to define and verify a formal group process for developing simulation model inputs when data is lacking.The recommended process modifies a Delphi process and employs a panel of subject area experts to provide estimates through several rounds of web-based surveys. After each round, the panelists' responses are analyzed, and a summary of the responses and comments from the previous round, as well as any supplemental information, is provided to the panelists to help them develop estimates in the next survey round. By sharing information, the panelists gain insight into the beliefs and opinions of their colleagues, resulting in a growing consensus about the questions addressed in the study.This process was implemented in the Colorectal Cancer Simulation Study to develop inputs for the simulation. As in this study and many other medical simulations, a number of inputs are unknown or uncertain because appropriate data does not exist or experiments cannot be performed to define the unknown inputs because of their grave nature. In this study, fifteen experts from the areas of gastroenterology, epidemiology, and microbiology were recruited to serve on the expert panel. Three rounds of webbased surveys were conducted to reach consensus on four different study objectives related to adenoma development and cancer progression. The final simulation model inputs were developed using the estimates and the VisiFit distribution-fitting software.To examine the flexibility, usefulness, and acceptability of the process, the expert panelists and the study's Advisory Board were sent evaluation surveys asking specifically about the group process and the resulting inputs developed. The panelists felt the process was flexible, required a minimal time commitment, and the web-based surveys were easy to use. The group dynamics throughout the surveying process allowed everyone to share information without worrying about dominance or groupthink. The information available during the process to support estimate development was adequate from the perspective of the panelists. Both the panelists and the Advisory Board found the inputs developed via the process to be consistent with real-world cases of adenoma development and cancer progression. They also believed the input estimates were more accurate than what one individual or an informal group could have developed. Since the group process was fully executed and a growing consensus of the estimates produced the final simulation model inputs, the process was clearly feasible. The cost of the process was easily justified because of the limited methods currently available to otherwise gain this information for use in the Colorectal Cancer Simulation Study. Because this information represents the best estimates available to date, and considering there is limited data to support formal analysis, the inputs developed as a result of this process are extremely valuable. The method developed was therefore deemed a success and contributes a method for developing inputs in the absence of data to the field of computer simulation modeling.