El 30/04/15 a las 23:29, Allin Cottrell escribió:
On Thu, 30 Apr 2015, Ignacio Diaz-Emparanza wrote:
> [...]
> Thank you, Allin.
>
> When you have time and if it is not excesively time-consuming, it
> would be good to add an option for importing the "Linear series" from
> TRAMO/SEATS results. This is the series with outlier and missing
> values corrected. (We could then consider 'Tramolin' package
> deprecated).
OK, that's now in CVS, along with a linearize() function which takes a
series as input and returns the TRAMO linearization (not yet
documented). These things may need some tweaking, but I've checked
against Tramolin and they seem to produce the same output -- except
when the sample range is restricted before calling the functions, but
I think there's a bug in Tramolin for that case. Try comparing:
<hansl>
include tramolin.gfn
open data9-3.gdt
smpl 1979:3 ;
series price_l1 = tramolin(price)
series price_l2 = linearize(price)
print price price_l1 price_l2 --byobs
</hansl>
Allin
Yes, I can see it. (The use of 'firstobs()' in the definition of 'scalar
ausprim = $t1-firstobs(t)+1' may be the problem).
I have run some examples with the new code and I think it is working
very well. One more thing for leaving it perfect can be to allow to save
this linearized series when "Time series model only" (only TRAMO) is
selected. I like to run only TRAMO because it gives a very more detailed
information about what outliers we have and of what type are them.
Furthermore obviously you cannot apply SEATS for annual series, but you
can apply TRAMO for detecting outliers, so this feature will be also
useful for annual series. I admit that covering all possible cases may
be complicated, but at least we could have a way to recover the
linearized series for TRAMO default parameter RSA=3 (as tramolin does).
--
Ignacio Díaz-Emparanza
Departamento de Economía Aplicada III (Econometría y Estadística)
Universidad del País Vasco, UPV/EHU
Tfno: (+34) 94 601 3732
http://www.ehu.eus/ea3