2024‐03‐19

Discussed prolonged MV model with Matt

• We probably need complex statistical models to include the gradient change in the variables. This may end up being a shift in methods focus, and may not be a good use of time.
• Instead, Matt suggests building 6 models, for day 1 data only, day 2 data only, and so on, then using the earliest model which has reasonable predictive performance.
• I’m going to try that. In parallel, Matt and I will look into how others handle gradient changes.