The LMSR method for providing a multidimensional understanding of growth standard in human fetuses.

A new nonlinear multivariate regression method called the LMSR method is proposed, by which a multidimensional understanding for the development process of human fetuses can be provided. Statistically important quantities such as median, skewness, coefficient of variation, and correlation of underlying structure can be described by corresponding smooth curves. Those curves can be obtained by a fine combination of a multivariate power transformation of data and penalized likelihood. It will be shown that the LMSR method and some associated tools are clearly efficient in analyzing development process of human fetuses.