what is an acceptable values for fval in solver?

Hello!!.

Can someone help me with this please.

When using a solver, example lsqnonlin, in your opinion, what is an acceptable values for fval in solver as 4.6744e-11?.

Thank you!!

[Juan Angel]

or, is it acceptable 1.1099e-09?

[Walter Roberson]

sqrt(1.1099e-09) corresponds to 3.3e-5. Is an error of 3 1/2 cents per $1000 in the forecast acceptable?

[Juan Angel]

Thank you very much @Walter Roberson, very good example!.

Answer 1

[Star Strider]

That depends entirely on the data and the model. The lsqnonlin function does not return an ‘fval’ output, however it does return the squared norm of the residual as the second output.

[Juan Angel]

Thank you very much Star Strider for your help, very kind!.

My code is a macroeconomic model where I simulate, for example, the productivity of the economy. Do you think 4.6744e-11 or 1.1099e-09 is acceptable as a squared norm? Thank you very much!

[Torsten]

As Star Strider already said: The absolute value 4.6744e-11 or 1.1099e-09 doesn't suffice to tell you whether the fit is good or bad. Plot empirical data against model data. Make statistical tests about the goodness of fit.

[Juan Angel]

Thank you @Torsten and @Star Strider.

[Star Strider]

Our pleasure!