A statistical method for extracting smallsignal FET models is presented. The method is derived as a maximum likelihood estimate of the model parameters with respect to measured S-parameters. The method accounts for uncertainties both in measurements and in extracted parasitic elements. An advantage with the proposed method is that the covariance of the estimate is obtained. The covariance is then used to calculate statistical bounds of the extracted parameters. Monte-Carlo simulations are used to verify the statistical bounds of the parameters. Good agreement between theory and experiments is obtained; especially for the parasitic elements.