1: \begin{abstract}

2: This paper is concerned with screening features in ultrahigh dimensional data analysis, which

3: has become increasingly  important in diverse scientific fields.

4: We develop a sure independence screening procedure based on the

5: distance correlation (DC-SIS, for short).

6: %The distance correlation was proposed  by \cite{Szekely:Rizzo:Bakirov:2007}

7: %for measuring dependence between two random vectors.

8: The DC-SIS can be implemented as easily as the

9: sure independence screening procedure based on the Pearson correlation (SIS, for short)

10: proposed by \cite{Fan:Lv:2008}.

11: However, the DC-SIS can significantly improve the SIS. \cite{Fan:Lv:2008} established the sure screening

12: property for the SIS based on linear models, but the sure screening property is valid

13: for the DC-SIS under more general settings including linear models.

14: Furthermore, the implementation of the DC-SIS does not require model specification (e.g., linear model or

15: generalized linear model) for responses or predictors. This is a very appealing

16: property in ultrahigh dimensional data analysis. Moreover, the DC-SIS can be used directly to screen

17: grouped predictor variables and for multivariate response variables. We establish the sure screening property

18: for the DC-SIS, and conduct simulations to examine its finite sample performance.

19: Numerical comparison indicates that the DC-SIS performs much better than the SIS in various

20: models. We also illustrate the DC-SIS through a real data example.

21: \end{abstract}

22: