Test for the Presence of Autocorrelation in the Buchanan model used in the Fitting of the Growth of the Catechol-degrading Candida parapsilopsis

Abstract

Catechol is a metabolic byproduct of phenol degradation by microbes. Its toxicity to human, mammals, insects and fishes has been long studied and its presence in the environment at toxic concentrations has been demonstrated. Fortunately there are microbes that could degrade catechol and can be used in bioremediation works. The growth of these microbes usually exhibit sigmoidal pattern due to the toxicity of the substrate. Previously, using the least square method in nonlinear regression, we report that the Buchanan three-phase model is the best model in fitting the growth of the yeast Candida parapsilopsis on this substrate. The ordinary least squares method relies heavily on several important assumptions such as residuals conformation to normal distribution, does not have outliers, is truly random, of equal variance (homoscedastic) and does not show autocorrelation. If all of these assumptions are satisfied, the test is said to be robust. In this work we perform statistical diagnosis test to test for the presence of autocorrelation as the growth model is time-dependent and many time-dependent curves shows evidence of autocorrelation.