The aim of the present study was to determine the physical properties of the Guazuma crinita wood by conventional methods, as well as to estimate these properties by near infrared spectroscopy (NIR) through the development of multivariate calibration models. The samples were obtained from five trees and from three levels longitudinally for 5 and 8 years of age randomly selected from a plantation in Puerto Inca, Department of Huánuco (Perú). Basic density, volumetric, tangential and radial contraction were determined by the conventional methods. Models were developed by regression of partial least squares (PLS) through mathematical transformations. For the validation of the models, a new group of samples -which did not participate in the development of the model were used. Statistical significance was found (F5 years = 191,5; F8 years = 84,9 ) between the different heights in the stem. Basic density and volumet ric contraction show values of Rc2 (0,88 and 0,83) and RPDp (1,55 and 3,27) acceptable for prediction.