Possible bias in analytical precision between the ICP and PXRF datasets, for the same Carrock Fell rock specimens, was assessed by linear regression. In this analysis the ICP data was used as the independent variable, which ideally should show zero error (Ramsey et al. 1995). Parametric hypothesis testing (F tests, Student t-tests, z tests) were used to investigate statistically (at the 95% confidence level) if the collection of implements was geochemically similar to the Carrock Fell Gabbro. Before testing, all datasets were investigated and found to be approximately normally distributed, thereby allowing the use of parametric tests. The level of geochemical similarity between individual stone implements and the Carrock Fell Gabbro was assessed by considering the proportion of elements measured that plot within the range of the means ± 1 standard deviation (SD) and ± 2 SD) of the derived Carrock Fell geochemical signature (Davis 1999).
© Internet Archaeology/Author(s) URL: http://intarch.ac.uk/journal/issue26/9/4.html
Last updated: Wed Jun 10 2009