Bibliometric databases on publications and their citations offer the possibility to quantify the scale and impact of this key output of scientific research. But there are many decisions to be made about how to construct these bibliometric measures, and no global consensus on how best to do so. We consider several measures of research output across scientific disciplines in New Zealand and test their sensitivity to a number of methodological choices. Factors considered include which publications to include in the analysis, ways to count publications co-authored by both local and international authors, and procedures for normalising citations. We also explore the potential to assign publications to disciplines by semantic analysis and a way to construct confidence intervals for the measures. The findings provide insight into the extent to which these methodological considerations impact on bibliometric results, and their interpretation, and how uncertainty in the results can be quantified.