Abstract
This study conducted covariance structure analysis in order to quantify the relationship among factors which caused field-to-field variability in 'Tanbaguro' production, using the data obtained by field investigations for 3 villages of Kyoto prefecture in 2007 and 2008. Firstly, we developed a structure equation model including 26 observed variables and 8 latent variables. The observed variables were consisted of soil chemical properties, weather condition and crop growth. The model explained 68% of sample covariance matrix which consists of covariance among observed variables. However, result suggested that the model was unstable because of too many factors. Accordingly, we improved the model by reducing the factors on the basis of causal coefficients. The improved model obtained 14 observed variables and 3 latent variables, and explained 79% of the sample covariance matrix. The improved model also suggests that early growth is important to obtain higher yield, and that yield and grain size are closely related with nitrogen status in a plant which defined by nodule condition and leaf color. The effect of water should be incorporated to the model in the next step.
