ABSTRACT

In the main body of this book, we aim to survey various hypothesis testing methodologies for functional data analysis. In the development of these methodologies, we essentially assume that continuous functional data are available or can be evaluated at any desired resolution. In practice, however, the observed functional data are discrete, probably with large measurement errors as indicated by the curve data sets presented in the previous chapter. To overcome this difficulty, in this chapter, we briefly review four well-known nonparametric smoothing techniques for a single curve. These smoothing techniques may allow us to achieve the following goals:

• To reconstruct the individual functions in a real functional data set so that any reconstructed individual function can be evaluated at any desired resolution.