Trees compete for space in the canopy, but where and how individuals or thier componnt parts win or lose is pooly understood. We developed a stochastic model of three-dimensional dynamics in canopies using a hierarchical Bayesian framework, and analysis 267533 positive height changes from 1.25 m pixels using data from airborne LiDAR within 43 ha on the widward flank of Mauna Kea. Model selection indicates a strong resident advantage, with 97.9% of positions in the canopy retained by their occupants over 2 years. The remaining 2.1% were lost to a neighborring contender. Absolue height was a poor predictor of success, but short stature greatly raised the risk of of being overtopped. Growth in the canopy was exponentially ditributed with a scaling parameter of 0.518. These findings show how size and spatial proxmity influence the outcome of competition for space, and provide a general framework for the analysis of canopy dynamics.…… かくのごとく, 三次元林冠データ,林冠動態 (消失と成長),階層ベイズモデル …… といった, いかにも私なんかが好きそうなかんじのものではあるのだが …… あまりちゃんと読んでないせいかもしれないけれど, いまいち面白くないんだよね. 何をやっているのかよくわからない統計モデリングのせいかな ……
library(R2WinBUGS)
利用の検討.
すでに WinBUGS が吐き出した init とかを読みこんで何とかしろ,
といったハナシ
……
bugs()
関数よびだしはこんなところか?
library(R2WinBUGS) post <- bugs( data = "data.txt", inits = c("inits1.txt", "inits2.txt", "init3.txt", "inits4.txt"), parameters = c( 'd.A.B', 'd.A.C', 'd.A.D', # 'w.A.B.C', 'w.A.B.C.D', 'w.B.C.D', 'sd.d' #, 'sd.w' ), model.file = "model.txt", n.chains = 4, n.iter = 8000, n.burnin = 5000, n.thin = 1, codaPkg = FALSE, # ディレクトリ指定など,あれこれ )