“…Successful modeling approaches involving wavelets and splines and adaptive kernels have been reported in the literature (Mohamed and Davatzikos, 2004; Morris and Carroll, 2006; Guo, 2002; Morris et al, 2011; Zhu et al, 2011; Rodriguez et al, 2009; Bigelow and Dunson, 2009; Reiss et al, 2005; Reiss and Ogden, 2008, 2010; Li et al, 2011; Hua et al, 2012; Yuan et al, 2014). A different direction of research has focused on principal component decompositions (Di et al, 2008; Crainiceanu et al, 2009; Aston et al, 2010; Staicu et al, 2010; Greven et al, 2010; Di et al, 2010; Zipunnikov et al, 2011b; Crainiceanu et al, 2011), which led to several applications to imaging data (Shinohara et al, 2011; Goldsmith et al, 2011; Zipunnikov et al, 2011a). However, the high dimensionality of new data sets, the inherent complexity of sampling designs and data collection, and the diversity of new technological measurements raise multiple challenges that are currently unaddressed.…”