Skip to contents

Squared Pearson Correlation (R2)

Usage

metric_R2(metric_dat, ...)

Arguments

metric_dat

dataframe with columns model and obvs

...

ignored

Details

Computes R-squared using the correlation-based formula: \(R^2 = \left(\frac{\sum(obs - \bar{obs})(mod - \bar{mod})} {\sqrt{\sum(obs - \bar{obs})^2} \cdot \sqrt{\sum(mod - \bar{mod})^2}}\right)^2\)

Note: Because this is a correlation-based R2, it is invariant to bias and slope. A model that is perfectly correlated but biased (e.g., model = obs + 100) will still score 1. This is distinct from variance explained or Nash-Sutcliffe Efficiency (NSE).

Author

Betsy Cowdery