Using data from 2,779 rabbits collected systematically from four sites in Australia, analyses have been made of the numbers of the nematode parasites Graphidium strigosum, Trichostrongylus retortaeformis, and Passalurus ambiguus. For statistical analyses simple logarithmic transformation was equally as efficient as transforming according to a fitted negative binomial distribution; in many cases a negative binomial did not satisfactorily fit the data. Using dried lens weight as an age-indicator the relationship of numbers of each parasite to host age was examined. A simple positive linear regression of numbers of Graphidium on lens weight was very highly significant. For Passalurus a quadratic relationship was revealed. No relationship was detected between numbers of T. retortaeformis and estimated host age. The implications of these age relationships are discussed. Following correction of the numbers of parasites to allow for the different host population age structures, it was possible to examine relationships with factors such as season and host sex. Analysis of the data for G. strigosunm revealed highly significant seasonal differences in abundance and even more significant between-sex differences-females harbored more parasites than males. At the most heavily infected and most studied site, the main feature with both T. retortaeformis and P. ambiguus was a highly significant sex times season interaction factor; during the breeding season females were much more heavily infected than males, a tendency towards the reverse being apparent in the nonbreeding period. The systematic collecting of a large number of wild rabbits, Oryctolagus cuniculus (L.), and the use of recently developed techniques for estimating their ages, has permitted a detailed analysis of the relationship between host age and sex, and season, and the numbers of parasitic nematodes. MATERIALS AND METHODS The data used for these analyses are basically those from which earlier papers (Dunsmore, 1966a, b, c) were prepared; the four sites and collecting techniques are described there. Data are also available from several further samples collected at the subalpine Snowy Plains site. Two smaller samples, included in the Snowy Plains analysis, were collected from a nearby and climatically similar site. Samples were collected as follows: Snowy Plains (subalpine): 25 samples totaling 1,324 rabbits. Urana (temperate): 9 samples totaling 481 rabbits. Mitchell (subtropical): 10 samples totaling 516 rabbits. Tero Creek (semiarid): 9 samples totaling 458 rabbits. The numbers of each nematode species were Received for publication 19 December 1967. * Division of Wildlife Research, CSIRO, Canberra, A.C.T., Australia. t Division of Mathematical Statistics, CSIRO, Canberra, A.C.T., Australia. estimated by counting all present in /25 of the material; we have used as the variable for testing goodness of fit to the negative binomial the observed number of parasites in this subsample. Bartlett (1947) discussed the necessity to change the scale of measurement of biological data in order to increase or establish the validity of statistical analyses-if the variance tends to change with the mean level of the measurements, the variance will only be stabilized by a suitable change of scale. If, in our data, a relation existed between means and variances, transformations should be considered; accordingly this possibility was first examined. A series of transformations was suggested by Bartlett for use with different distributions of observed data and accordingly an attempt was made with the present data to judge the efficiency of the commonly used logarithmic transformation as compared with a transformation based on a fitted negative binomial distribution; several authors (e.g., Northam and Rocha (1958), Froyd and Clarke (1962), Williams (1964), Thomas (19651)) have suggested this as the distribution followed by numbers of metazoan parasites. The method used for fitting the negative binomial was that used by Pahl (in press). The Bliss and Fisher (1953) method of estimating k is not suitable for computer programming; therefore the Newton-Raphson method for determining the maximum likelihood estimate for k was used (Pahl, pers. comm.). The efficiency of the transformations was judged by the degree of fit of the transformed variate to normality, the extent to which the variances were stabilized between different samples and over a range of means, and, rather more empirically, the discriminant ability of transformed variables in subsequent analyses.