成長解析のための補助的手法としてのBayes型モデルの応用  [in Japanese] An application of a Bayesian model as an auxiliary method for analyzing growth  [in Japanese]

成長速度の変動を表すモデルを探索するための補助的手法について展開した。成長係数の時間変化は一般に特定の周期関数でモデル化されるが,成長係数の1次階差を確率的に記述する方法を用いた。パラメータ推定とモデル選択は,Bayesの定理と周辺尤度に基づいた。従来の手法との比較のため,サクラマスの体長データを解析した。成長メカニズムの情報が不十分なとき,データに基づくモデルの探索が必要である。 Bayes型モデルの補助的利用により,データのもつ特性をより詳細に把握できると思われる。

Before constructing growth formulae with time-varying growth coefficient, the gain in body length should be examined using data. For this purpose, it is necessary to apply an auxiliary analysis model while constraining the fluctuation of body expansion in time. This paper presents an analysis model which has a flexible fitness for the fluctuation in length. In general, the time-varying growth coefficient has been described by a periodic function. 0n the contrary, we employed the first-order differences of growth coefficient, which follows the probability distribution. Here, we describe the parameter estimation and model selection based on Bayes' Theorem and Marginal Likelihood. In order to compare the result of the Bayesian approach with that given by the conventional way, we introduced the body length data of masu salmon. In a case where only unreliable growth information has been obtained, it is necessary to extract the biological characters from data, which is useful for model exploration and construction. Through the auxiliary utilization of the Bayesian model, the biological features related to the complicated mechanism of growth may be specified.