Mh eval

[mhgov]

Evaluation notebooks

Still has some outstanding actions for Alex/Andreas to pick up when required (bold)

For topic modelling (Alex)
PS: We want to understand the variation in approval rate for different topics so we can focus are efforts to add content, revise script on specific
Steps:

1. Locate data of historical messages / responses in download from s3 (done)
2. Use topic modelling to categorise the message / responses (done)
3. Look at the approval rate of the messages (done)
   4) Use the router instead of topic modelling and recreate same analysis (to do)

For rct evaluation (Andreas)
PS: Need to prep a dataset and a notebook in case of an FOI request
Steps:

• Speak to andreas and see if he has the original analysis

1. Download the results (done)
2. Find only the RCT results (partially)
   **3. Filter for RCT results we trust
3. Graph the survey responses
4. Save graphs as .pdfs and required anon data into .xlsx for sharing**