model_error_measurement: Model error measurement metrics

Description Usage Arguments Details Author(s) References

Description

Common interface for metrics commonly for assessing quality of hydrologic models, statistical or otherwise.

Usage

1
2
3
4
5
excelR2(x.obs, x.model)
rmse(x.obs, x.model)
rmse(residuals)
nash.sutcliffe(x.obs, x.model)
nash.sutcliffe(x.obs, x.model, x.alt)

Arguments

x.obs

- observed values

x.model

- modeled values

x.alt

- alternate model for comparison (Nash Sutcliffe only, defaults to mean of observed)

Details

excelR2 returns the R^2 as reported by Excel's curve fits. It is provided as people are comfortable with it, but is a terrible measure of model accuracy. See the references below.

rmse returns the root mean square error. With one parameter assumes x.obs is residuals, with two, computes residuals between x.obs and x.model

nash.sutcliffe returns the Nash-Sutcliffe model coefficent, greater than 0 if 'x.model' is a better fit than x.alt, 1 if a perfect fit, and between 0 and -infinity if a worse fit than x.alt.

Author(s)

Evan Heisman

References

Hopper, T. (2014), Can We do Better than R-squared? http://www.r-bloggers.com/can-we-do-better-than-r-squared/ retrieved 16 May 2014

Nash, J. E. and J. V. Sutcliffe (1970), River flow forecasting through conceptual models part I - A discussion of principles, Journal of Hydrology, 10 (3), 282-290.


eheisman/hydroutils documentation built on May 13, 2017, 3:02 p.m.