A New Method to Predict Seed Yield of Moist-Soil Plants

Abstract

Multiple linear regression can be used to predict seed yield of moist-soil plants; however, measurement of multiple predictor variables is tedious, subject to variation, and these models can exhibit multi-collinearity. Thus, we tested if simple linear regression models could predict seed yield of 5 species of moist-soil plants as precisely as multiple linear regression models. The single predictor variable was number of dots on a grid covered by seed. Simple regression models explained as much variation in seed mass (R 2 adj = 0.92-0.97) and predicted (R 2 pred = 0.91-0.96) as well as or better than multiple regression models. Precision of models was attributed to the strong positive linear relation between the dependent variable and the predictor, accurate dot counting, and lack of multicollinearity. Dot counting also was easier and more efficient than measuring multiple phytomorphological variables. This new method is useful for researchers and managers estimating seed yield of moist-soil plants; however, additional models should be developed for other plant species, and the method should be tested in other regions.