Speaker
Andrew Nelson
(ANSTO)
Description
refnx [1] is a next generation reflectometry analysis package, building on its predecessor, Motofit. In this presentation we discuss its main design features:
- Bayesian statistics core with comprehensive uncertainty analyses and model selection ("how many layers can the data justify") [2].
- modular construction of structural models, ranging from a basic Slab up to freeform SLD profiles and Lipid membrane leaflets. These components are easily extensible.
- Co-refinement of multiple contrast datasets.
- Mixed Area models.
- Python based with analyses performed in Jupyter notebooks or a Qt GUI.
refnx is specifically designed to facilitate reproducible research. Here we also discuss what reproducible research means in the context of a neutron scattering study, outlining how this is achieved with refnx, and how these practices could (should) be taken up by neutron scatterers in general.
[1] Nelson, A. R. J. & Prescott, S. W.
refnx: neutron and X-ray reflectometry analysis in Python
Journal of Applied Crystallography, 2019, 52, 193-200
[2] If you don't know what Bayesian statistics is and have always wondered, you'll now find out.
| Level of Expertise | Expert |
|---|---|
| Do you wish to take part in the poster slam | No |
Primary authors
Andrew Nelson
(ANSTO)
Stuart Prescott
(UNSW Chemical Engineering)
