Premium
Combined Analysis of Yield Data from Fertilizer Experiments 1
Author(s) -
Laird R.J.,
Cady F.B.
Publication year - 1969
Publication title -
agronomy journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.752
H-Index - 131
eISSN - 1435-0645
pISSN - 0002-1962
DOI - 10.2134/agronj1969.00021962006100060001x
Subject(s) - regression analysis , yield (engineering) , mathematics , statistics , regression , standard error , variables , variable (mathematics) , linear regression , zea mays , interaction , partial least squares regression , econometrics , agronomy , biology , physics , mathematical analysis , thermodynamics
The corn ( Zea mays L.) yield data obtained with four levels of nitrogen at 76 locations were combined to give a general equation with yield expressed as a function of applied N variables, productivity factors, and interaction variables. Three regression procedures — stepwise, backward elimination, and an approach based on agronomic considerations — were employed in producing a general yield equation. The pooled blocks within sites error and combined experimental error, calculated from the analyses of the individual experiments, were used for testing the significance of the site variable and plot variable regression coefficients, respectively. High correlations existing between pairs of the independent variables prevented a meaningful interpretation of the coefficients in the estimated regression models. The total variation in yield accounted for by the three regressions was similar, although the terms dropped from the full model in producing the agronomic model resulted in a significant reduction in the regression sum of squares. However, when the estimated models were used to predict yields for sites not included in the estimation procedure, the agronomic model predicted better than the other models.