Model data structure planning

[emmamendelsohn]

Assemble all data sources into time series

• Each row is a pixel for a date from 2006 (start of WAHIS reporting) through present. Select a random day from each month.
  • Use the finest resolution pixel available from the data
• Trailing 90 days of recorded weather data aggregated by month
• Forecasted 90 days and a weight based on time since last seasonal forecast was released
  • Use average and variance of forecast models (iterations)
• Variable(s) indicating if there has been an outbreak in last month(s) in pixel, adjacent pixel, distance weighted pixel, in country
• Variable(s) indicating frequency of outbreaks in previous years, which could increase immunity, and whether there were vaccinations or other controls in response to outbreaks
  • Need to pull control data from wahis
• Outcome variable should not include wildlife, but past and adjacent outbreaks can include wildlife

[emmamendelsohn]

combined these notes with issue #24