1: \begin{abstract}

2: Sliced inverse regression

3: (Duan and Li [\textit{Ann. Statist.} \textbf{19} (1991) 505--530],

4: Li [\textit{J. Amer. Statist. Assoc.} \textbf{86} (1991) 316--342])

5: is an appealing dimension

6: reduction method for regression models with

7: multivariate covariates. It has been extended by

8: Ferr{\'e} and Yao [\textit{Statistics} \textbf{37} (2003) 475--488,

9: \textit{Statist. Sinica} \textbf{15} (2005) 665--683]

10: and Hsing and Ren [\textit{Ann. Statist.} \textbf{37} (2009) 726--755]

11: to functional covariates where

12: the whole trajectories of random functional covariates are completely

13: observed. The focus of this paper is to develop sliced inverse

14: regression for

15: intermittently and sparsely measured longitudinal covariates.

16: %Previous work in dimension reduction with functional covariates (

17: %trajectories of random functional covariates are observed completely,

18: %but our approach allows for longitudinal covariates that are recorded

19: %discretely, intermittently, and maybe sparsely.

20: We develop asymptotic theory for the

21: new procedure and show, under some regularity conditions, that the

22: estimated directions attain the optimal rate of convergence.

23: Simulation studies and data analysis are also provided to demonstrate

24: the performance of our method.

25: \end{abstract}