1: \begin{abstract}

2: Shape constraints offer compelling advantages in nonparametric

3: regression by enabling the estimation of regression functions under

4: realistic assumptions, devoid of tuning parameters.  

5: However, most existing shape-constrained nonparametric regression

6: methods, except additive models, impose too few restrictions on the

7: regression functions.  This often leads to suboptimal performance,

8: such as overfitting, in multivariate contexts due to the curse of

9: dimensionality.  On the other hand, additive shape-constrained models

10: are sometimes too restrictive because they fail to capture

11: interactions among the covariates. In this paper, we introduce a novel

12: approach for multivariate shape-constrained nonparametric regression,

13: which allows interactions without suffering from the curse of

14: dimensionality. Our approach is based on the notion of total concavity

15: originally due to T. Popoviciu and  recently described in

16: \citet{gal2010shape}.  We discuss the characterization and computation

17: of the least squares estimator over the class of totally concave

18: functions and derive rates of convergence under standard assumptions.

19: The rates of convergence depend on the number of covariates only

20: logarithmically, and the estimator, therefore, is guaranteed to avoid

21: the curse of dimensionality to some extent.   We demonstrate that

22: total concavity can be justified for many real-world examples and

23: validate the efficacy of our approach through empirical studies on

24: various real-world datasets.     

25: \vspace{0.5em} \\

26: \noindent\textbf{MSC 2010 subject classifications:} Primary 62G08. \\

27: \noindent\textbf{Keywords and phrases:} Interaction effect modeling, mixed partial derivatives, multivariate convex regression, nonnegative least squares, Popoviciu's convex functions, shape-constrained estimation. 

28: \end{abstract}

29: