Downloads & Free Reading Options - Results

We extend nonparametric generalized linear models to allow both the mean curve and the response distribution to be nonparametric. The seemingly intractable task of working with two infinite-dimensional parameters is shown to be reducible to a finite optimization problem, which is easily implemented via existing algorithms. We demonstrate using various examples that the proposed approach can be a flexible tool for data analysis in its own right, but can also be useful for model selection and diagnosis in a more classical generalized linear model framework.