I am currently modelizing the price of real-estate in France. I chose three type of models : an arima, an mce, a var ... i of course fitted them in the best way possible thourgh the usual prodecure (box-jenkins and such) ... NOW

I am (killing myself trying) to identify and index that would allow me to select the best model.

Currently reading the Model Selection and Multimodel Inference: A Practical Information- Theoretic Approach, Second Edition by Burnham & Anderson

but i get the feeling that this metric (AIC) is used to select "a model" ... well this is the issue actually. I am not trying to select the best model amongst (ie) different regressions (same instrument) ... but the best model between different instruments (arima, var, mce)

I of course can identify it graphically but i need to have a metric ?

Currently researching the AIC, BIC, HQ, RMSE, MAE, MAPE and studying the general Kullback et Leibler distance

but i have an issue whenever the litteratre states " selecting amongst differents models " ... because i have to select among different "instruments"

NB : i KNOW i'm not making any sense right now.


where can i find intel on formulas and how to perform, MdAPE, MdRAE and RMSE in R ? i found a meta study detailing the performance of error measurement techniques on wiki (year 1992) http://faculty.weatherhead.case.edu/Fred-Collop/researchArticles/ErrorMeasures.pdf

(chap 8 for results, chap 3 for definition of perf. criterias)

| cite | improve this answer | |

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.