The effects of three error sources in plasma curve measurements on parameter estimation in kinetic analysis of positron emission tomography (PET) fluorodeoxyglucose (FDG) data are investigated by computer simulation. The three error sources are: (1) measurement noise in the radioactivity concentrations of plasma samples; (2) linear interpolation between adjacent plasma sampling points of the plasma time activity curve; and (3) incorrect weights used for the least-squares regression. All three error sources are found to increase the variability of the parameter estimates, with the first one a primary error source in normal PET FDG studies. The performance of five estimation methods which account for the error sources are evaluated. When the noise variances of the plasma and the tissue measurements are not known, an iterative weighting procedure is shown to give accurate and reliable estimates.