Speaker
Description
Reproducibility is a broad term that covers many aspects of the scientific endeavour. In the context of a reflectometry experiment, ‘reproducibility’ can be understood to cover the details of sample preparation, data collection, data reduction, and data analysis. Attention to reproducibility is enabling: it allows the experiment to be repeated, the data handling pipeline to be checked, the data to be reanalysed, and the model to be reused.
As data is passed from instrument to instrument scientists to user to journal article it is also transformed from complex formats with rich metadata into lowest-common-denominator formats often with no metadata. While ‘standard’, the data transformations are seldom published and the low fidelity of the transforms with regards metadata frustrates reproducibility. Significant progress has been made in the reproducibility of data analysis, with more publications making use of open source code that has been described in the literature; the specific details of the models including code to reproduce the analysis is now routinely featured in the publications from some groups.
Recent achievements in improving the reproducibility of NR analysis will be illustrated using details from selected publications. Examples of good practice will be discussed, along with challenges (caused by current workflows, software, and file formats) that either prevent or discourage reproducibility.
