We consider a situation where the distribution of a random variable is being estimated by the empirical distribution of noisy measurements of that variable.
This paper studies linear panel regression models in which the unobserved error term is an unknown smooth function of two-way unobserved fixed effects.