Pharmacokinetic parameters estimated by nonlinear weighted least one multiplication.doc

 PAGE \* MERGEFORMAT 16 Pharmacokinetic parameters estimated by nonlinear weighted least one multiplication Author: Lee Jin-Wen Sun Yan, Zhao-Hui Chen Zeng-ping Abstract presented pharmacokinetic (PK) parameter estimation of non-linear weighted least one multiplication (ML1). The meaning of the objective function so that the weighted residuals and the absolute or relative mean absolute error (AARE) tends to the minimum. Using genetic algorithms (GA) on the nature of the objective function almost no requirements, easy to get the global optimization parameter estimates, and the sound is good. Application of GA to estimate PK model parameters, the effect better. Keywords: pharmacokinetic parameters of the estimated weight of the robustness of Genetic Algorithm In the current literature [1 ~ 4], pharmacokinetics (PK) model fit when the objective function Q is generally: minQ = σWiε2i (i = 1,2,3, ..., n) (1) Type (1), Q for the residual or weighted residual sum of squares; Wi-weighted (usually 1,1 / Ci or 1/C2i, etc.); εi that the drug concentration measured for the residual value and the corresponding PK model ( parameters θj, j = 1,2,3, ..., p, p the number of independent parameters) Calculate the value or the difference between the predicted values, εi independent and identically distributed, εi expectation that the mean zero, variance of each point εi are equal to or approximately equal to σ2 (this time without weights or may consider it necessary to weight, but Wi are 1) or not equal, respectively σ2i (required at this time weighted); n for the sample size that the drug (C)  time (t) the number of data. According to nonlinear least squares (LS) and the principle of optimality, so that Q tends to the minimum, you have desires PK parameters of the LS estimate of θ. In general, C  t curve of a larger difference between high and low concentrations can vary several times, a few times or even hundreds of times, in most ca