On the estimation of functional random effects Articles uri icon

publication date

  • February 2017

start page

  • 50

end page

  • 58

volume

  • 17

international standard serial number (ISSN)

  • 1471-082X

electronic international standard serial number (EISSN)

  • 1477-0342

abstract

  • Functional regression modelling has become one of the most vibrant areas of research in the last years. This discussion provides some alternative approaches to one of the key issues of functional data analysis: the basis representation of curves, and in particular, of functional random effects. First, we propose the estimation of functional principal components by penalizing the norm, and as an alternative, we provide an efficient and unified approach based on B-spline basis and quadratic penalties.

keywords

  • functional pca; subject-specific curves; identifiability; sop algorithm; models; components; regression; principal; curves