Hello,

You can deal (as alternative of GARCH) with EGARCH model to avoid the assumption of non-negativity of parameters. Juts a note, why you didn't adjust the IVs in the mean equation rather than variance -equation (or both of them)!!

Chellai F

The model equation is in order as it is. The parameter values of the independent variables can be negative or positive.

Chellai Fatih I considered changing to a EGARCH model and putting some of my IVs in the mean equation as you suggested me. I kept cpi (corruption perception index) and gov(government integrity) inside the variance eq because I consider these variables affect more the variance than the mean of my dependent variable (usd/cop fx). Here are my new results, any comment that helps me with the interpretation and analysis?

Ette Etuk The problem is that I don't know if this parameters will affect the positivity of the variance, do you know about any documentation that supports that the parameter values of the independent variables inside the variance equation can be negative or positive? thanks for your help

Isabel Venegas , ( I think you konw already the compenents of an EGARCH model); so , for mean equation the interpretation is directely, rest for the variance equation , in your last fitted model,

c8: is simply the ARCH effect (or spilover effect)

c9: is the asymmetry coefficient (it is postiive (0.0768) and statisticaly significant) whcih means that good and bad news of the same level have different impact on the volatility)

c10: measure the persistence in volatiltiy (past volatilty explain the current volatility as in AR models).

Chellai f

Your model has loggarch as the subject of the formula rather than garch. Therefore the coefficients can be negative.