Mixed effects regression models in forestry research

Abstract

A promising method for finding patterns in experimental data is regression models of mixed effects, which have not found wide application in forest science in Russia to date. Over the past two decades, interest in them has grown significantly in the world scientific community. Mixed-effect models are extensions of regression models for data that is collected by group. In forestry, for example, individual stands, trial plots, geographic regions, etc. can be considered as separate groups influencing the resultant trait. Compared to classical fixed effect models, the addition of a random component avoids violating the assumption of independence in repeated measurements. Therefore, parameter estimates are more reliable. Mixed-effect models are used to solve a wide range of problems in forestry, from describing pairwise relationships between individual tree variables to reflecting the dynamics of forest stands. By giving more accurate forecasts of variables in comparison with traditional models, which include only fixed effects, their introduction into production activities can increase labor productivity and economic efficiency of forestry. A large positive experience of using models of mixed effects abroad should not go unnoticed in the domestic forestry science. Their active use makes it possible to reveal new or hidden patterns in experimental data, thereby giving a new vector in the development of forestry, forest inventory and other forestry scientific disciplines.

Keywords

• models of mixed effects
• fixed effect
• random effect
• tree stand
• forestry