Data
For dugong growth, the observations CSV tells you which lengths were used to fit the programs and which were held out for scoring. Open the predictive-curves CSV to read each program's median prediction at each input and the 5th to 95th percentile band around it. The program summary gives coverage by a central 90% predictive interval: how many of the 20 held-out observations lie between the 5th and 95th percentiles, including any on the bounds.
Dugong growth
- Held-out predictions (CSV)
- Predictive curves (CSV)
- Program summary with scores and coverage (CSV)
- Four-chain fit data (CSV)
- Observations (CSV)
Disease spread
- Held-out predictions (CSV)
- Predictive curves (CSV)
- Program summary with scores and coverage (CSV)
- Four-chain fit data (CSV)
- Observations (CSV)
In the held-out prediction files, heldout_log_score is the total log score of a program over its 20 held-out observations and avg_heldout_log_score is that total divided by 20; both are repeated on every row for that program.