Analyzing Categorical Soil Test Information.

Many soil test procedures are used for P and K determination in North America, although just a few are dominant. In order for data to be pooled among laboratories using different procedures, ranges of agronomic equivalency for each test were defined. These ranges were either taken from the literature or estimated by soil fertility specialists in consultation with IPNI North American or Nutrient Program Directors. All soil test data were reported in terms of well known soil test procedures. Maximum likelihood estimation (MLE) was used to generate parameters of four distribution models: lognormal, gamma, Weibull, and Burr. The distribution model with the lowest Akaike information criterion (AIC) was considered to “best” fit the observed categorical data. Changes over time were determined using logistic regression, analyzing each soil test category separately and using the sample numbers from each year as weights.