The Robustness of TOPSIS Results Using Sensitivity Analysis Based on Weight Tuning

Abstract

Multiple-criteria decision analysis (MCDA) is one of the support techniques for Health Technology Assessment (HTA). A typical method is the Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS). The output of TOPSIS presents an order of alternatives. A significant risk may be posed by an inappropriate setup of the weights of the criteria such that a very small deviation from the proper value could substantially change the final result. Sensitivity analysis is a method for testing changes of the final order by a modification of the original input data or by small deviation of the original weights of the criteria. This original approach is very slow and computationally demanding in the case of a change of the value of any attribute. The newly proposed method is significantly faster, making it possible to change of the values of the weights by only a few computations within TOPSIS. In the first stage, TOPSIS is used to compute the values of the Positive and Negative Ideal Solutions (PIS and NIS), and sensitivity analysis is only performed for the changed weights in the next stage. In the proposed method, a weight is adjusted (and the other weights recalculated) in order to find a range where the order of the alternatives remains unchanged for the modified weights. The weight adjustment can have the form of a fixed change or of an iterative process approximating the stability range for the weight. The new method is fast and simple to implement, providing a robust output of TOPSIS and pointing out a possible wrong setup of the initial weight(s).