Gidas: Estimation and Consistency for Optimal Transformations for 
       Prediction in Continuous-Time Processes

NOT REALLY AN ABSTRACT       
The talk is about a nonparametric estimation (and consistency)
problem in time series with continuous-time parameter. "Optimal
Transformations for Prediction" is a pair of NON-TRIVIAL functionals
of the "past" and "future" that minimize the squared error. Both 
existence and estimation/consistency are delicate problems; in the
talk I will emphasize the estimation and consistency problem.