Acceptable range for the goodness of fit

Hello every one,

I have estimated sets of data that should be compared to the reference data. I used the MATLAB code of 'goodnessOfFit' with the NRMSE and I got these values: -0.1716 and 0.3712. I want to know what is the acceptable value range? and which one is better?

Is there any more underestandable way to know the goodness of fit?

[Walter Roberson]

For this purpose we need to know if you were using R2020a or later, or if you are using an earlier version of MATLAB.

[Star Strider]

I would rely more on the confidence intervals for the parameters and the fitted curve, or other derived statistics (such as p-values).

[Mitra Taghizadeh]

@Walter Roberson

Hi Walter,

I'm using 2019b.

[Mitra Taghizadeh]

@Star Strider

Hi

Just for clarification I should say some more explenation.

Actually, I have two sets of experimental data for displacement and I have fitted two curves on them using cftool. Now obtained one graph by combination of these two obtained graphs and I want to know how much this new graph is fitted to the experimental data.

Do you think that I should use P-value? How can I use it?

[Star Strider]

The p-value should be less than 0.05 conventionally to determine significance.

The confidence intervals would be more revealing, because they easily determine if a parameter is significantly different from zero (both confidence intervals of the same sign).

If you are comparing two different models with the same number of parameters, use the Likelihood-ratio test to determine which is best.

If the two models have different numbers of parameters, this gets complicated quickly. See your Statistics Department for guidance in that event.

[Mitra Taghizadeh]

Thank you.

[Star Strider]

My pleasure!