Error in r sq log

unread,

Feb 7, 2022, 11:44:19 AM2/7/22

to hydromad users

G'day, can someone tell me why the value of r sq log is wrong:
rel bias: -0.2446
r squared: 0.8566
r sq sqrt: 0.623
r sq log: -0.1514

unread,

Feb 7, 2022, 3:53:06 PM2/7/22

to hydromad users

Hi,

It's not clear to me why you think it's wrong, but the equation is on the help page:

Do note that there is an offset to allow it to be applied to zero values.

Cheers,

Joseph

unread,

Feb 15, 2022, 12:07:54 PM2/15/22

to hydromad users

Thank you for your reply, but I want to know is that normal to have a negative value of r sq log?

greetings

unread,

Feb 15, 2022, 4:02:58 PM2/15/22

to hydromad users

Hi,

The Nash Sutcliffe Efficiency family of objective functions compares performance to a benchmark; R squared and r sq log use the mean.

A negative value means that the model is doing worse than simply using the mean as the estimate.

Your model looks like it has overfit to r.squared and is performing poorly in terms of r sq log (and not great on rel bias either)