Diagnose weather support for a prediction
Description
Counts weather terms outside their observed training ranges and reports the largest absolute standardized anomaly from the training mean. This diagnostic is marginal: it does not establish support of joint weather combinations, probability of error, or valid prediction interval coverage.
Usage
prediction_support(object, newdata)
Arguments
object |
A fitted 'crop_yield_model', including an ensemble. |
newdata |
Prediction panel or 'crop_yield_data' object. |
Value
A data frame with region/year, 'OutsideTerms', 'TotalTerms', 'MaxAbsZ', and 'UnseenRegion'. Ranges are pooled across training regions.
Examples
data <- synthetic_crop_example()
model <- fit_crop_model(data, "trend")
prediction_support(model, subset(data$panel, Year == 2016))
[Package yieldlag version 0.3.0 ]