Data Analysis 1a(i) Minutes
- Brian Maranville (chair,scribe)
- Ben Ocko
- Mark Schlossman
- Wojciech Potrzebowski
- Thomas Saerbeck
- Gaetano Mangiapia
- Juan Manuel Carmona Loaiza
- Jean-Francois Moulin
There is a desire to have the calculation engine separable from the graphical interface.
Wojciech - We can look to other parallel efforts offering recommendations in the area of software development for scattering and reflectometry, in particular the outputs of the SINE2020 working group
Ben - Construction of flexible platform would be useful, where people can put in their own algorithms.
Ben - Noted that there is a large software group at his facility, and that we may need to engage with the people actually writing the software in order to make progress on some of our goals
later in the conversation it was noted that Juan is actually one of the developers of BornAgain, and was on the call
Mark - there are programs/utilities available at his facility that do reading, and analysis of the software, handling building of a model by elemental composition. They are addressing multiple types of analysis on the same sample, such as fluorescence and reflectivity. (using Certified Scientific Software, a company: https://certif.com)
A general note that very few programs for reflectivity analysis have scripting capabilities for large numbers of datasets.
Gaetano - kinetics measurements in reflectivity can produce 100 - 200 datasets, and need to be simultaneously fitted to a shared set of parameters across files.
There were some questions about BornAgain, and how they handle the model-building for specular reflectivity, whether it is a three-dimensional model as is required for the off-specular and GISANS modeling also available in that package. I think the answer was that it is a layer-based model system separate from the 3d models, and that there is a model-builder built into the program.
Ben - really wants to include a way to add weighting to datasets so that higher-Q data gets more emphasis in the fits, which can be dominated by the low-Q data (much smaller relative error bars). He does not use Chi-Squared as the main goodness-of-fit evaluation
Wojciech - pointed out that there are tutorials available for BornAgain, which is very helpful for getting users engaged. This overlaps with Education and Outreach mission.
- identify developers actually working on data analysis software (and engage with them if not already done)
- add scripting capabilities to list of desired functionality in reflectometry analysis packages
- see how dissimilar inputs are for some common analsysis packages (Refl1D and BornAgain as a start)
- survey users/scientists to get an idea of what everyone is using for analysis
- add soft X-ray analysis as a desired functionality