Distinguishing between the confounding effects of temporal dependence, variation in exogenous factors and residual heterogeneity over and above that due to measured explanatory variables is a major challenge to be confronted in any analysis of panel or similar longitudinal data. This paper addresses the main issue that arises in this context, that of controlling for residual heterogeneity, and reviews two nonparametric methods that have been proposed. These methods are of some practical interest because of evidence that longitudinal models are not always robust to alternative parametric specifications of the residual heterogeneity. Their use is illustrated by three examples, covering residential mobility, depression and unemployment. The compirical results also demonstrate some of the misleading consequences of failure to account for residual heterogeneity. Alltention is drawn to the computational and other problems which appear to have inhibited the adoption of nonparametric control.