[Gretl-users] Re: testing for Long Memory using Fractional Integration test

Am 05.12.2019 um 09:33 schrieb Yusuf Abduwahab Hassan: > Good morning, Please i want to investigate the integration properties > of a variable so as to analyse the persistence in the series. 1. i > want to test the fractional integration hypotheses within the > interval (0,1) with grid ie, with step, 0.1. Is my procedure below > correct? >> fractint ld_ELCNigeria(0) --all No, there are two problems here: - The numeric optional parameter to the 'fractint' command must not be enclosed in brackets, but with blanks, like this: fractint ld_ELCNigeria 0 --all - But this is not what you want, either, because the null hypothesis of the test is always I(0), as stated in the help text. Instead the numeric parameter AFAICS refers to a lag (truncation) order, not the integration order. (It is true that this is formulated a little bit confusingly in the help text and this probably should be changed.) So if you want to run a test for I(0.1) you would have to fractionally difference the series first, with d=0.1, and then run the test above. Gretl has the 'fracdiff()' function to do that kind of differencing. So the following should work (untested): series temp = fracdiff(ld_ELCNigeria, 0.1) fractint temp --all You can also manually specify the (lag) order if you like, but I cannot tell you which value to choose for that. > 2. how can i employ the Gil-Alana(2002) code in Gretl -- Please provide a link to the code you have in mind, or if it is short, maybe you can also paste it into the mail message. cheers sven _______________________________________________ Gretl-users mailing list -- gretl-users(a)gretlml.univpm.it To unsubscribe send an email to gretl-users-leave(a)gretlml.univpm.it Website: https://gretlml.univpm.it/postorius/lists/gretl-users.gretlml.univpm.it/