Statistical Analysis of Population Density Data of Soil Animals, with Particular Reference to Enchytraeidae (Oligochaeta)

Abstract

The fitting of the negative binomial distribution to populations of enchytraeid worms collected in coniferous forest soil is investigated. A regression between the two parameters m and k of the negative binomial distribution has been established. This regression is compared with the result of similar studies on other soil animals. This comparison shows that the variation in k is very great, but the regression seems to describe the average relationship between x and k found for the other soil animals. The effect of this non-normality upon the application of parametric, statistical methods used to compare means is examined using simulation in an electronic computer. It is found that the confidence intervals do not cover the true mean as often as expected according to the normal assumption. The discrepancy between the observed and the expected confidence probability is, however, not very significant when the mean is larger than about 10. A logarithmic transformation of the observations has been found to distort severely the confidence probability. However, with regard to the probability of type-I errors of the analysis of variance, the transformation has had a slight improving effect. The analysis of variance has, however, been found robust against the non-normality of the original data. Transformation for this purpose seems therefore to be unnecessary.