Identification and diagnostic tests for ARCH

ARCH models are almost always estimated using the maximum likelihood method
and with the use of computationally expensive techniques. Although linear GARCH
fits well with a variety of data series and is less sensitive to misspecification, others
like the ARCH-M requires that the full model be correctly specified. Thus identification
tests for ARCH effects need to be carried out before, and misspecification tests after
the estimation process are necessary for proper ARCH modeling.