I am busy validating a model for "run-off" using river flow information. I have approximately 5 years of daily data and have been looking at G-causes as an additional statistic to help explain the difference between observed and simulated run-off within a catchment.
Using R and running the granger.test() from Library(MSBVAR) I have the following results
granger.test(Daily.sea, p=10) F-statistic p-value Observation -> Simulation 3.363867 0.0002219447 Simulation -> Observation 44.726567 0.0000000000
Is it fair to say the both Observation g-causes Simulation and vice versa? This indicates that there is indeed a relationship between the two variables. I ask because my results using hydroGOF are not that good (as in weak relationships between the two).