Abstract
The performance of two median-based methods, namely, median method (MM) and least median of squares (LMS), for parameter estimation in Arrhenius equation is investigated and compared with least sum of squares, through extensive Monte Carlo tests involving simulated data. Reweighted methods associated with LMS are also included in the study. Different types of experimental noise are considered for simulating variable standard deviation (SD) of noise and the occurrence of outliers. The results indicate the relative precision and accuracy of various methods for each noise type, and hence are useful in selecting the most appropriate method for a specific application provided noise type is known. Often little or no information is available on variation of SD and occurrence of outliers, and in such situations the most appropriate method seems to be MM which is observed to be the best for many noise types.