1: \begin{abstract}

2: The problem of estimating the mean of random functions based on

3: discretely sampled data arises naturally in functional data analysis.

4: In this paper, we study optimal estimation of the mean function under

5: both common and independent designs. Minimax rates of convergence are

6: established and easily implementable rate-optimal estimators are

7: introduced. The analysis reveals interesting and different phase

8: transition phenomena in the two cases. Under the common design, the

9: sampling frequency solely determines the optimal rate of convergence

10: when it is relatively small and the sampling frequency has no effect on

11: the optimal rate when it is large. On the other hand, under the

12: independent design, the optimal rate of convergence is determined

13: jointly by the sampling frequency and the number of curves when the

14: sampling frequency is relatively small. When it is large, the sampling

15: frequency has no effect on the optimal rate. Another interesting

16: contrast between the two settings is that smoothing is necessary under

17: the independent design, while, somewhat surprisingly, it is not

18: essential under the common design.

19: %Furthermore, our results show that for sparsely sampled functions,

20: %the independent design leads to superior convergence rate

21: %when compared to the common design, and therefore should be

22: %preferred in practice.

23: \end{abstract}