A Statistical Comparison between Different Multicriteria Scaling and Weighting Combinations

Abstract

Multicriteria decision making presents several challenges for researchers. These include selecting a technique that will produce accurate results but not require too much time or resources. Researchers have been comparing different techniques for several years, though a comprehensive study of using many different techniques for the same set of problems is relatively new. Additionally, knowing if the difference in scores between alternatives is significant or not presents another challenge. The use of confidence intervals has recently been employed by researchers to examine whether results are actually statistically different from one another. This study uses 21 different scaling-weighting combinations and confidence intervals on six different decision-making problems to measure their ability to produce unambiguous results. Linear normalization as a scaling technique tends to be the best at identifying one or two clear winners while avoiding complete ambiguity. Despite the variety of combinations used, several common themes emerge across the decision problems.