The 5 best beta regression models (measured by Akaike’s Information Criterion corrected for small sample sizes [AICc]) to predict a participant’s score on the written exam from independent variables (type of training, city of training, participant’s previous experience with plants); models were based on pooled data from three sites. Model parameters include the number of predictors (K), AICc, comparison of model AICc with AICc of the best model (ΔAICc), goodness of fit (ModelLik), and log likelihood (LL).
| Model | K | AICc | ΔAICc | ModelLik | LL |
|---|---|---|---|---|---|
| Training | 3 | −121.02 | 0 | 1 | 63.80 |
| Intercept only (null) | 2 | −118.14 | 2.88 | 0.24 | 61.21 |
| City | 4 | −115.15 | 5.87 | 0.05 | 62.06 |
| Training + Experience | 7 | −112.05 | 8.97 | 0.01 | 64.50 |
| Experience | 6 | −108.72 | 12.30 | 0.00 | 61.44 |