Understanding how predictor changes drive outcomes with a flexible, nonparametric approach.
This work introduces the average derivative estimation (ADE) method for studying multivariable regression relationships. It shows how to estimate a key set of coefficients and then model the mean response as a nonparametric function of a weighted sum of predictors.
The book explains why ADE offers a practical alternative to fully parametric models. It combines simple, interpretable summaries of variable impacts with graphical tools that reveal nonlinearity. Using kernel methods and density estimation, it stays data-driven and avoids strong model assumptions while delivering dimension-aware insights.
- How average derivatives summarize the relative influence of each predictor on the mean response
- How to estimate these coefficients and the underlying regression function nonparametrically
- How ADE can reveal nonlinear patterns through plotting the ADE regression estimator
- Evidence from Monte Carlo simulations and comparisons with multivariate smoothing
Ideal for researchers and practitioners seeking a flexible, interpretable way to explore complex regression relationships without heavy parametric risk.