Tobit regression is a special kind of regression. For regression analysis goodness-of-fit between real and fitted data is usually tested by Pearson's chi-square statistical test or the G statistical test. These can equally be done with the Tobit regression.

What kind of goodness-of-fit statistic are you thinking about?

One simple possibility is to correlate the predicted values and actual values. This gives you a type of r-squared. An example of this is shown at the end of this page: https://stats.idre.ucla.edu/r/dae/tobit-models/

Both are correct, but because MLE is used for Tobit regression, we typically use log-likelihood measures for fit values. In general, including a fit measure such as Madalla R-Squared along with the comparison of predicted/actual values is the way to go. One could, however, also use McFadden's Pseudo R-Squared as a fit measure, being that MLE is used. AIC and BIC are also fit measures that can be used.

Your interpretation will be different depending on what fit measure you elect to use.

Thank you all for your suggestions