1: \begin{abstract}

2: \noindent The paper is concerned with asymptotic properties of the

3: principal components analysis of functional data. The currently

4: available results assume the existence of the fourth moment.

5: We develop analogous results in a setting which does not

6: require this assumption. Instead, we assume that the observed

7: functions are regularly varying. We derive the asymptotic distribution

8: of the sample covariance operator and of the sample functional principal

9: components. We obtain a number of results on the convergence of

10: moments and almost sure convergence. We apply the new theory

11: to establish the consistency of the regression operator in a functional

12: linear model.

13: 

14: \medskip

15: 

16: 

17: \noindent

18: Key words: Functional data, Principal components, Regular variation

19: 

20: \end{abstract}

21: