Uso de análise exploratória de dados e de regress?o robusta na avalia??o do crescimento de espécies comerciais de terra firme da Amaz?nia

摘要

This research originated from a data base of two forest inventories at 100%, carried out in 1984 and 2000 on an area of 576 hectares of an upland primary forest, located in the National Forest of Tapajós, in Belterra, Pará, Brazil. The objective of the research was to use exploratory data analysis and robust regression for modeling diameter and basal area growths. In the two inventories, circumference breast height (CBH)-cm, was measured with a common tape; the commercial heights were measured with a Weise hypsometer in 1984 and, estimated with the use of sticks in 2000. The minimum DBH was 55 cm in 1984 and 35 cm in 2000, for all species. The statistical analyses were preceded by exploratory data analysis (EDA), where the box plot was used for outliers detection, and the stem-and-leaf graph to filter the extreme observations. Robust regression was used to adjust the models in the outliers presence. This research showed that, in spite of the intra and inter specific variations the six species together presented an observed and expected growth rate in diameter and basal area of 4,56 and 4.28 mm/year and 13.00 and 13.09 cm2/ha/year, respectively. It was also shown that the use of exploratory data analysis and robust regression made possible the analysis and determination of periodic increments in diameter and basal area in a consistent way. The methodology used for the available data and variable types was found to be more efficient when using robust regression for the method Least Trimmed Square than for the ordinary minimum square method.