Table 2.
Analysis of selected published peer-reviewed studies and the relevant metrics used to assess model performance and their respective numerical values for the best and the worst candidate models.
| Study | Quality | Study Location | Study Design | Statistical Methods | Metric | Best Model Value | Worst Model Value |
|---|---|---|---|---|---|---|---|
| Donalisio (2008) | + | Brazil | Ecological | None | N/A | N/A | N/A |
| Donalisio (2011) | ++ | ENM | t-test | p-value | 0.02 | 0.0001 | |
| Prist (2016) | ++ | Bayesian model | t-test, one-way ANOVA, and Tukey’s MCT | p-value | 0.9231 | 0.0626 | |
| Prist (2017) | + | Bayesian model | Moran’s I, Queen’s case neighbourhood relation | SD of risk values | 3.4 | 3.8 | |
| Muyalert (2019) | ++ | Zero inflated model and temporal term (rw2) + Besag ICAR spatial model | INLA, Moran’s I, cross validation procedures | CPO | 0.58 * | 0.36 | |
| Andreo (2014) | ++ | Argentina | GLM and MaxEnt models | Kruskal-Wallis and pairwise Pearson correlation | AIC | 76.07 | 100.42 |
| Vadell (2019) | ++ | GLM | Kappa index, VIF analysis, bootstrap procedure, and residual plotsAugmented Dickey-Fuller Test, AICc, RMSE & Ljund-Box test | Kappa index | 0.58 | 0.44 | |
| Explained deviance | 70.5% | 61.2% | |||||
| Ferro (2020) | ++ | ARIMA & dynamic regression models | AICc | 97.74 | 291.4 | ||
| RMSE | 0.91 | 0.78 | |||||
| R2adj | 0.71 | 0.54 | |||||
| Montgomery (2012) | - | Bolivia | Epidemiology | None mentioned | N/A | N/A | N/A |
| Nsoesie (2014) | ++ | Chile | ARIMA and regression with ARIMA errors | Coefficient of variation R2 and RMSE | AICc | 437.62 | 444.98 |
| RMSE | 2.421 | 2.774 | |||||
| R2 | 0.594 | 0.551 | |||||
| Bayard (2004) | − | Panama | Outbreak investigation | None mentioned | N/A | N/A | N/A |
| Williams (1997) | − | Paraguay | Outbreak investigation | None mentioned | N/A | N/A | N/A |
| Douglas (2021) | + | Barbados | Cross-sectional epidemiology | 95% confidence intervals (CI) | N/A | N/A | N/A |
Key ++ very high quality study; + high quality study; − low quality study; N/A—not applicable; R2—r squared-coefficient of determination; R2adj—r squared adjusted is r squared that has been adjusted to the number of predictors in the model; *—This is an optimised mean to improve CPO values from a zero-truncated Poisson model; Akaike information criteria (AICc); Analysis of variance (ANOVA); Autoregressive Integrated Moving Average (ARIMA); Cross validation procedures (CPO); Ecological Niche model (ENM); Generalised linear model (GLM); Intrinsic Conditional Auto-Regressive (ICAR); Integrated nested Laplace approximations (INLA); Maximum entropy (MaxEnt); Tukey’s multiple comparison tests (MCT); Root mean square error (RMSE); Standard deviation (SD); Variance inflation factor (VIF).