Accounting for multiple effects and the problem of small sample sizes in osteology: a case study focussing on entheseal changes

Abstract

Osteoarchaeological studies provide valuable information concerning living conditions and life course changes in past societies. However, many skeletal markers, such as entheseal changes, are multifactorial in aetiology; thus, their interpretation is not straightforward. Generalised linear models (GLMs) are ideal for analysing such phenomena, i.e. those with multiple underlying causative factors, but, to date, their use has been limited. This paper focuses attention on using these models to test hypotheses regarding the aetiology of entheseal changes, widely regarded as indicative of activity patterns, but which are also affected by ageing and body size. To demonstrate the use and limitations of these models, this paper provides an independent test of a previously developed GLM on an identified skeletal sample comprised of skeletons from four British sites (n=58) which has a typical sample size for archaeological osteological analysis. In addition to this model, GLMs were developed to include the factor of body size and expand the models to test individual entheses, as well as joint complexes whereby multiple entheses for muscles which act synergistically have been pooled. The results indicate that the original model did not compare well with the frequencies of entheseal changes found in the British assemblage under study. The new models found no clear pattern of influence, although both ageing and body size were important for some entheses. GLMs are appropriate for testing the interaction of biological variables, but future studies need to take into account and test their applicability to archaeological sample sizes.

A Test of the Effectiveness of the Coimbra Method in Capturing Activity‐induced Entheseal Changes

“…Such methods allow the study of the impact of activity patterns on ECs while controlling for the effect of other contributing factors. The results have so far supported the primary role of age in EC expression, followed by body size, while activity patterns also appear to exert a subtle impact on the observed changes (for example, Henderson & Nikita, ; Michopoulou et al, , but see also Villotte et al, for a more pronounced impact of activity). Despite the availability of statistical tests that allow the effective study of multifactorial phenomena, such as ECs, the optimum way to record ECs in order to capture activity‐induced alterations remains unclear.…”

A Comparison of Two Methods for Recording Entheseal Change on a Post‐Medieval Urban Skeletal Collection from Aalst (Belgium)

“…As with most archaeological studies, it must be borne in mind that limited sample size will impact the statistical results (for a discussion of the effect of sample size on EC, see Henderson and Nikita ); however, as the sample sizes are highly similar for both methods (Table ), it can be assumed, at least for concordance between methods, that the impact of sample size was limited.…”

Application of geographic information systems to investigating associations between social status and burial location in medieval Trino Vercellese (Piedmont, Italy)

“…For the spatial analyses described here, we used z ‐scores to standardize osteometric values for each individual by sex and status subsample. z ‐scores are commonly employed to examine growth outcomes in human biology and have been introduced in bioarchaeology to expand sample size by combining populations subsamples (e.g., Pinhasi, Timpson, Thomas, & Slaus, ; Zakrzewski, ), to examine skeletal parameters of different absolute dimensions concomitantly (e.g., Henderson & Nikita, ; Schrader, ), and to allow for direct comparisons of samples from different points in time and space (Farley, Moll, & Blacksher, ; Vercellotti et al, ). z ‐scores standardize data by the mean and standard deviation of a normal distribution.…”