Analysis of variance, inference, multiple comparisons and sampling effects in soil research

The analysis of variance is a crucial step in extracting information from efficiently designed experiments and surveys in soil science. It is only the beginning, however. From it, follow the standard errors (SEs) of means, SEs of differences and other effects provided by experiments, which in turn lead to tests of significance. Use the simple least significant difference (LSD) at some acceptable probability for testing comparisons of individual means. Do not use experiment-wise multiple comparison tests. In experiments with graded treatments do not make multiple comparisons of any kind; instead fit a response curve and analyse the data by regression. Sampling fluctuation within experimental units and surveys contributes short-range variation to the residual variance of measured soil properties and increases errors. Diminish this contribution either by replicate sampling and measurement within plots or by bulking before measurement. Sample all replicates in the field; sampling in the laboratory (pseudo-replication) is no substitute. In almost all investigations the mean values for experimental treatments and survey classes are the most important outcomes. So report them with their SEs; readers will then be able to make of them what they will.