ESTIMATION OF TRUE BIOMASS ENERGETIC YIELD AND MAINTENANCE COEFFICIENT WITH ERRORS IN THE VARIABLES

Simple regression analysis is based on the least square of residual er rors. To use this method, an independent variable which is observable without error is assumed to exist. For most analyses, the assumption i s normally made without realising the implications. The present study demonstrates how parameters may be estimated with the assumption of er ror in all measurements. Data sets involving batch/fed-batch cultivati on of Candida utilis were analysed using a statistical approach with e rror in variables for the estimation of true biomass energetic yield a nd maintenance coefficient. This approach requires replicated measurem ents. For linear systems, such as in chemical/biochemical processes, r eplicated measurements are sometimes possible or there are physical/ph ysiological reasons to group the observations into clusters. When clus ters of replicated measurements are not available, the cluster analysi s technique may be used to identify groups of near-replicated data poi nts. The procedure yielded similar results to those obtained by the or dinary least squares procedure for consistent data. However, for data of poor quality (especially data with large errors in the independent variable), the disparity between the two procedures was more significa nt.