ABSTRACT
Pon~el. Properties of Gootincss-of-Fit Tests 347
TABLE 21.1 Proportion of rejections in 100 samples of the hypothesis of white noise
h1 particular, and followirig Heltriio and Bloomfield (1987) we select I r as the nlinilnixer of the function
The rcsult,s of Table 21.1 corifirrn olu. a.syir~pt,otic i~~~irlysis. As it is seen the tests maintair~ the level and the tcst outperforms the iilT,IL lest for q - l . Recall thitt for this value of cl the d rp ;~r t i~re from the 111111 is caused by a nonvanishing first order irutocorrel:~liorl. However, the power ol' thc test clrops off' clrastic;ally for. the valucs y = 3 arid (1 - 6 since Sor these valurs of q the d e l ~ i ~ ~ l l l l ~ c f wn tlie inlll is clue to ii. sigriificmt al~tocorrelatioi~ at higher lags, c.g., lag 3 a d Ci ~.espectivcly. For these values of q the i";T,h t,cst clearly outperforms the (+ test wncl its power tli.ops more gri~(ltl~tlly iis q irlcrxasrs.
