Problem with garch (tseries)

Se?ores
Can i do some forescast of log returns ???
i?m doing  an ARMA-GARCH process  step by step 
rt=ut+et
 
why am i do this?? because just like this i can to remove the no significative parameters and come back to estimate it?s that right??, (because it  exist  the great fSeries package, and there?s no reason for no use it???)



now can i predict return of this way....???


predict(arma)+ predict(garch) of any sentence of the times series packages????


thanks a lot 

pd: by the way ... can i use the iid.test package to seek the behavior of the returns????  

Ricardo Zambrano Aguilera
Chile