Methods for estimating parameters of Holling's disk equation (=Michaelis–Menten function) were investigated mathematically and statistically. The sensitivity of each method to changes in variances was examined with computer simulation. The results showed that when error variances are proportional to the number of prey attacked, the Woolf transformation is effective in both linearizing the disk equation and reducing correlations between the variance and the mean, thus yielding the best estimates. When error variance is constant, nonlinear regression provides the best estimates. Under both circumstances, estimates by the proportional I (Holling's proportional method) and proportional II transformations are biased, because of correlated error terms. Estimates by the Lineweaver–Burk transformation have the least accuracy. Reanalysis of 22 data sets from the literature revealed that the Woolf transformation yielded better parameter estimates than nonlinear regression in 13 cases. Nonlinear regression yielded better estimates in two cases and tended to overestimate one of the parameters. The estimates of the other seven cases were similar between the two methods. We recommend that the Woolf transformation be used for estimating parameters of the Holling's disk equation instead of nonlinear regression.