testing of statistical significance of parameters which are under the restriction, base on T test can be correct? if not, how can I make inference about estimated parameter?
An example of a parameter which is under constraint is the standard deviation just like every other measure of dispersion. Dispersion is a distance and therefore cannot be negative. The test of significance of a standard deviation can be based on the sampling distribution of the variance which is chi-square. In garch modelling alpha is constrained to be less than 1 for stationarity to be attained and positive for positive residual variance. In spite of this constraint, statistical significance of the alpha coefficient might be tested by a t-test of the mean value (within the limits of the constraint). Therefore the use of the t-distribution for the test of significance of a constrained parameter is acceptable where applicable.