Provide feedback on SACRO test script

[lucyb]

Note: This issue should be moved into the SACRO repo once we have one

Note: This issue may need further information, so ask Lucy or Seb, if required

❗ Deadline is end of May

Lucy/Seb needs to provide feedback on our experiences using the SACRO test script to Jim at UWE. This ideally needs a researcher to be involved.

The process to follow is:

1. The latest version of the code is on pipit so can be installed using: pip install acro

2. We’ve taken the decision to use an open-source well documented dataset of data about nursery admissions. It is fully described and available for download from here: https://www.openml.org/search?type=data&sort=runs&id=26&status=active

3. We've written an example script for people to try - it is described in this notebook: which you can view online. It doesn’t display all of the functionality yet, and we haven't finished translating the R version to this dataset (but can if you really like) https://github.com/AI-SDC/ACRO/blob/main/notebooks/test-nursery.ipynb

[sebbacon]

Things to consider:

• The most important feedback UWE needs is "this doesn't work" or "this doesn't make sense" - high level testing and sense-checking
• The ideal outcome is that we eventually use this as decision support tool for our own output checking, so feedback should also have this in mind. "Will this support your own output checking process? If not, why not? What would you change?"
• We can use the testers' experience with the tool to inform how we incorporate its outputs into the concept mockups

[bloodearnest]

I've dumped some initial thoughts from trying there demo code here https://bennettoxford.slack.com/archives/C63UXGB8E/p1684409451900489

[benbc]

@bloodearnest When you get back to this (maybe after you're back from holiday), can you go through it again with @HelenCEBM so we can include her thoughts?

[bloodearnest]

https://docs.google.com/document/d/15ElDrxxyBhfByJfPuN1Dkc-qc_0vpv8mmF3-9DbxM5E/edit