I am trying to find the best model to predict the measured values. I have 8 models. To compare between the models to find the best one, I am trying the best fit equation, standard deviation, Mean, Coefficient of variation and Taylor diagram for each model. Can you please suggest other ways of comparison to find the best model ?
You should explain a bit more. With that info at hand I would say maybe BIC, AIC, or BFs.
There is a family of measures sometimes grouped as measures of "accuracy". Each is based on the observed values, predicted values, and potentially the difference between these two called residuals. Some are in the original units of the observed variable and some are normalized to a unit-less measures. These include mean absolute error, median absolute error, mean absolute percent error, root mean square error ,normalized root mean square error, Efron's pseudo r-squared, and coefficient of variation.
Another group is based on the likelihood from the model. These include AIC, AICc, BIC, and pseudo r-square measures like McFadden, Cox and Snell, and Nagelkerke.
Which if any of these makes sense depends on the type of model and the nature of the observed values.
- Badges
- Science topic
- Similar topics
- Filing