Re: [Gretl-devel] grabbing Alfred

Am 21.08.2013 09:40, schrieb Sven Schreiber: I'm attaching a hansl file with a couple of relevant functions which should cover many real-world cases. These functions assume an Alfred csv file preprocessed like I described before, which resides in the location specified in the string argument 'fname'. (This path is taken as-is, so the caller is responsible for getting it right.) These are the function signatures, where 'vname' is the string variable holding the variable name as in the Alfred file: 1) getAlf_ithPub(fname,vname,i[0::1]) Retrieves for each observation period (= what I have also called reference period sometimes) the i-th published value, no matter when that publication occurred. Can lead to NAs, e.g. for the future or for high publication numbers. The case i=0 is special, it gets the final publication for each obs period. Examples: series @vname_latest = getAlf_ithPub(fname,vname,0) series @vname_second = getAlf_ithPub(fname,vname,2) 2) getAlfVintage(fname,vname,day) Retrieves the vintage valid on date 'day' (which is a string argument in ISO format), for all obs periods. This is perhaps the classic meaning of realtime data series. If the vintage date is the present, then this should coincide with the latest publication: getAlf_ithPub(fname,vname,0). Will also lead to NAs, e.g. for obs periods that are younger than the vintage date. Example: series @vname_19830302 = getAlfVintage(fname,vname,"1983-03-02") 3) getAlfPubafterN(fname,vname,n) Retrieves for each obs period the publication which was current as of (the first day of) the n-th period (quarter or month) after the respective obs period. So in some sense it imposes a constant publication lag of n periods, no matter how many publications occur in between. Also possible in principle is n=0, for variables where the first publication occurse instantaneously (financial markets?). OTOH, can lead to NAs especially for low n. Example: series @vname_after2 = getAlfPubafterN(fname,vname,2) In all these cases, the dating (=time axis) in the gretl workfile is according to observation or reference periods. The publication time of each value differs in general and depends on the context in a relatively complicated way. Of course, much more testing needed. cheers, sven