Merge pull request #84 from epiforecasts/text-update

small rephrase

Author: kathsherratt

report/results.Rmd

@@ -127,7 +127,7 @@ effects_comp <- results$effects |>
   select(-c("effect", "Unadjusted"))
 ```
 
-Descriptively, we noted apparently similar predictive performance between mechanistic, semi-mechanistic, and statistical models. These model structures appeared to perform relatively worse than agent-based and "other" models. For example, in univariate analysis, the partial effect for statistical models forecasting deaths underperformed by `r table_effects["Deaths_Unadjusted_Statistical","value"]`, (95%CI `r table_effects["Deaths_Unadjusted_Statistical", "lower_2.5"]`-`r table_effects["Deaths_Unadjusted_Statistical", "upper_97.5"]`) compared to average, while agent-based models performed better than average (`r table_effects["Deaths_Unadjusted_Agent-based","value_ci"]`). However, variation in performance overlapped between all model structures, and we noted relative differences between models may have varied over time (Figure \@ref(fig:scores-over-time)). For example, over summer 2021 all model types saw worsening performance coinciding with the introduction of the Delta variant across Europe, but this decline was most marked among statistical models of death outcomes compared to any other model type.
+Descriptively, we noted apparently similar predictive performance between mechanistic, semi-mechanistic, and statistical models. These model structures appeared to perform relatively worse than agent-based and "other" models. For example, in univariate analysis, the partial effect for statistical models forecasting deaths indicates underperformance by `r table_effects["Deaths_Unadjusted_Statistical","value"]`, (95%CI `r table_effects["Deaths_Unadjusted_Statistical", "lower_2.5"]`-`r table_effects["Deaths_Unadjusted_Statistical", "upper_97.5"]`) compared to average, while agent-based models performed better than average (`r table_effects["Deaths_Unadjusted_Agent-based","value_ci"]`). However, variation in performance overlapped between all model structures, and we noted relative differences between models may have varied over time (Figure \@ref(fig:scores-over-time)). For example, over summer 2021 all model types saw worsening performance coinciding with the introduction of the Delta variant across Europe, but this decline was most marked among statistical models of death outcomes compared to any other model type.
 
 These differences between model structures largely disappeared after adjustment for covariates. We found no clear evidence that any one type of model consistently outperformed others (Figure \@ref(fig:plot-coeffs)). There was no difference in accuracy between model structures when predicting cases, and we observed only weak differences when predicting deaths. In contrast to unadjusted estimates, we identified that statistical models may have performed slightly better (partial effect `r table_effects["Deaths_Adjusted_Statistical","value_ci"]`), and semi-mechanistic models worse (`r table_effects["Deaths_Adjusted_Semi-mechanistic","value_ci"]`) than the average, although with overlapping uncertainty.