Comparison of Ordinary Least Square and Mixed-Effect Regression Models for Estimation of Tree Diameter Increment: A Case Study for Dipterocarpacea in Siem Reap, Cambodia

• Main Authors: Suriyati Harun, Yasmin Yahya, Nurashikin Saaludin, Wan Suriyani Che Wan Ahmad, (UniKL MIIT)
• Published: ACM 2015
• Subjects: Ordinary least square model; mixed effect model; repeated measures; model validation
• Online Access: http://localhost/xmlui/handle/123456789/9719

Diameter increment is one of the common and important tree characteristics used in forest management decision making. In this paper, we compared different statistical methods to develop diameter increment models for individual tree of dipterocarpaceae tree species in semi-evergreen forest in Seam Reap, Cambodia. The two chosen methods were ordinary least square (OLS) and linear mixed effect model (LME) techniques. The aim was to clarify the effect of taking into account the repeated measurement of the data. Instead of the predicted parameters, we focused on the performance between the methods. The models were validated in terms of mean square error and mean percent error using independent test data set. The result showed mixed effect model was superior as it produced smaller prediction error.