DATA-ANALYSIS STRATEGIES IN FOREST RESEAR CH - DATABOX AND MULTIVARIATE CROSSPRODUCT-PARTITIONING (MCP) WITH WOOD SCIENCE SERVING AS AN EXAMPLE

The inhomogeneity of wood as a raw and working material must be accept ed as one of the basic empirical facts of wood science (GRAMMEL 1989; HOLZ 1965). Different influencing factors are responsible for this, e. g. growth conditions or the position of a sample in the tree. With the Multivariate Crossproduct-Partitioning (MCP) or Databox Partitioning described in this paper a statistical concept for differential wood sc ience is proposed, whose central issue is the description, explanation , prediction and control of wood properties in relation to growth cond itions. An example from a wood science project by SAUTER (1992) serves to illustrate this strategy. The transfer of this formal concept to o ther branches of forest science appears both sensible and feasible.