Statistics in soil science

Key points • Soil varies from place to place; statistics is a technology for describing that variation quantitatively and separating signal from noise. • Replicate sampling is required to estimate the variation in experimental material and geographic regions. • Randomization is needed to estimate mean values and their variances, and to place confidence limits on means. • Data can be summarized by means, medians, variances, standard deviations and skewness coefficients. • Distributions of data are best illustrated by histograms and Q–Q plots. • Analysis of variance apportions the variance in a set of data to sources in the designs of experiments and surveys; any actual analysis must fit the design by which the data have been obtained. • Relations among two or more measured variables are expressed by Pearson correlation coefficients. • A variable may be predicted from one or more others on which it depends by regression.