spell sensitive

inst/create-metric-tables.R

@@ -285,7 +285,7 @@ log_score <- list(
   where $p_y$ is the probability assigned to outcome p by the forecast F.
 
   **Usage and caveats**:
-  Larger values are better, but sometimes the sign is reversed. The log score is ensitive to outliers, as individual negative log score contributions quickly can become very large if the event falls in the tails of the predictive distribution, where $f(y)$ (or $p_y$) is close to zero. Whether or not that is desirable depends ont the application. In scoringutils, the log score cannot be used for integer-valued forecasts, as the implementation requires a predictive density. In contrast to the crps, the log score is a local scoring rule: it's value only depends only on the probability that was assigned to the actual outcome. This property may be desirable for inferential purposes, for example in a Bayesian context (Winkler et al., 1996). In settings where forecasts inform decision making, it may be more appropriate to score forecasts based on the entire predictive distribution.)"
+  Larger values are better, but sometimes the sign is reversed. The log score is sensitive to outliers, as individual negative log score contributions quickly can become very large if the event falls in the tails of the predictive distribution, where $f(y)$ (or $p_y$) is close to zero. Whether or not that is desirable depends ont the application. In scoringutils, the log score cannot be used for integer-valued forecasts, as the implementation requires a predictive density. In contrast to the crps, the log score is a local scoring rule: it's value only depends only on the probability that was assigned to the actual outcome. This property may be desirable for inferential purposes, for example in a Bayesian context (Winkler et al., 1996). In settings where forecasts inform decision making, it may be more appropriate to score forecasts based on the entire predictive distribution.)"
 )
 
 wis <- list(