Abstract
A random-effect model is devised with individual- and time-specific components for uncensored data, based on a specification for censored data developed by Anderson (Citation1985) and extended by Kim and Maddala (Citation1992). In addition, specification tests are devised that allow one to choose between the model and a conventional random-effects model.
Acknowledgements
Financial support from FONDECYT Grant No. 1030887, and from an institutional grant of the Hewlett Foundation to CEA is greatly acknowledged. All remaining errors are the author's.
Notes
1 Kim and Maddala also consider a multiplicative heteroscedastic specification due to Anderson (Citation1985), in which for i = j, t = s, i, j = 1, … , N; s, t = 1, … , T; 0, otherwise.