COMMENT A SHORT NOTE ON THE PERFORMANCE OF TWO COMPUTER PROGRAMS FOR THE ESTIMATION OF THE PARAMETERS OF A MULTI-EXPONENTIAL MODEL

During the fitting of a multi-exponential model to experimental visco-elastic stress-decrease curves determined after stepwise straining of urinary bladder-wall tissue, Cool-saet et al. [l] noted a discrepancy between the results obtained with the two computer programs used to estimate the parameters of the multi-exponential model (i.e. the coeffi-cients and exponents of the various exponential terms). A sophisticated program called Exposum [Z], based on a Marquardt iteration, was used initially [3]. It was later found [l] that a lower sum of squares, indicating a better fit, was obtained with the program Expostep described by van Mastrigt [43, which is based on a simple stepwise exploration of the parameter space. Kuik [S] made the following remarks concerning this dis-crepancy in results: 1. If Exposum is used in double precision for analysis of the urological data, it gives the same sum of squares as Expostep or even a slightly lower one (owing to the fact that the parameters are determined at a higher precision). 2 The minimum in the sum of squares found in fitting these urological stress-decrease curl,‘es is situated in a long straight trough parallel to one of the parameter axes, prob