Dear everybody which is the best way to obtain a Goodness of fit in generalized least squares statistic analisys?
Here (http://avansp.github.io/2014/11/02/DIC-AIC-BIC/) is discussion of some of these. Lots of people use other methods, like types of cross-validation or chi-sq deviance. Given the debate I think many readers will say there is no "best way" for all situations, but it depends on your situation and needs.
Hello. The goodness of fit is usually judged by a measure on the residual error. In least squares this can be evaluated using the estimation error covariance. This usually requires the second order noise statistics to be known. You are probably aware of the overfitting problem - you can reduce the error by adding degrees of freedom. So judging the best "fit" is also tied up with choosing the minimum number of free parameters to get a "parsimonious" model.
The best measure of he goodness of fit is the coefficient of correlation between the observed values of the dependent variable and the fitted values of the dependent variable. It is as simple as that!
Thanks a lot to everybody. In terms of publication what is really needed to be written in the manuscript when I present my GLS results?
The coefficient of correlation between the actual and the fitted values of the dependent variable. Isn't that obvious?
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression