Methods for Estimating Parameters
After we decide on the data to be modeled, then we have to decide how to estimate the model parameters. Three methods are commonly used. The weighted least-squares method estimates parameters that minimize the weighted sum of squared differences between the model predictions and observed data. The maximum likelihood method estimates parameter values that maximize the likelihood of the observed data according to the model. The Bayesian method is different than both of the above because rather than computing a single point estimate, it computes an estimate of the entire probability distribution of the parameters given the observed data according to the model. Below we illustrate each method using the PVL model applied to choice data from the IGT.