INVESTIGATION OF THE ERROR STRUCTURE OF THE CALIBRATION CURVE FOR PERIODATE DETERMINATION BY FLOW-INJECTION ANALYSIS AND CHEMILUMINESCENCE DETECTION

Experimental data on periodate determination by flow-injection analysis and chemiluminescence detection is fitted to a model equation using the pattern search method. A large number of parameter estimates are obtained from simulated data generated by means of a computer program implemented by a random number generator, and their parameter estimate means and variances are compared with the set of parameters obtained from the experimental data. The %-bias, %-excess variance, skewness, and excess kurtosis, of the parameter estimates from the simulation study are also calculated to determine the effect of errors on the parameter values. This procedure was followed for different (a) criteria of convergence of the fitting routine, and (b) type of errors in the response value (constant/relative) and outliers selected at random. The statistical parameters of the model equation parameter estimates are compared in relation to the various convergence criteria for best performance. The results are discussed and the conclusions are criticized in order to explore improvements of the treatment.