Re: [Gretl-users] glitch with X-12

Allin: The latest version does seem to have resolved the "mini-error box" problem I was having. Now, though, there seems to be a minor gnuplot issue if I don't want the graphs. John: Many thanks for your advice. I have had a brief look at some of the X-12 output (I'm definitely in the "most users won't read this..." camp!). The trading day effects don't seem to be significant, even in the case when I get a warning about "visually significant spectra" (see #2 below). The survey on which this data is based always refers to employment/unemployment during the week containing the 12th of the month, which could maybe induce a trading day effect (??). Also, the automatic procedure is selecting additive adjustment, which is what I want. Here's the brief experiment I conducted: 1) E_T with trading day adjustment -> E1.txt. No errors or warnings about trading day effects, but I do get a gnuplot error if I uncheck the "generate graph" box: "C:\Program Files\gretl\wgnuplot.exe" "C:\Documents and Settings\pesummer\Application Data\gretl\gpttmp.a06048": exit code 1" 2) U with trading day adjustment -> U1.txt. "WARNING: At least one visually significant trading day peak has been found in one or more of the estimated spectra." However, the trading day effects seem to be insignificant in the X12 output. 3) E_T as before (with trading day adj.) -> E2.txt. This step was where I was getting the "mini-error" box, but that's gone. I've attached the data set and X-12 output files. Bottom line: I'm getting what I think are sensible results with the new default settings, apart from the gnuplot errors. PS -----Original Message----- From: gretl-users-bounces(a)lists.wfu.edu [mailto:gretl-users-bounces@lists.wfu.edu] On Behalf Of John C Frain Sent: Sunday, October 10, 2010 1:40 PM To: Gretl list Subject: Re: [Gretl-users] glitch with X-12 Allin Thanks. The new system should produce reasonable seasonal adjusment in a large number of cases. In Peter's case the problem is that the data may need some farther examination. This is an X12arima rather than a gretl question. Something of the order of 30% of series may need further attention. For an unemployment series I would expect that the adjustment would be additive rather than multiplicative. One expects unemployment to rise of fall with the season by numbers of persons rather than by a percentage of the current numbers unemployed. I would not expect a trading day effect if the measurement was always taken on the same day of the week each month. (If the measurement was taken on the last day of the month one might get a trading day effect). My advice to Peter, for the cases which appear to have a trading day effect, would be to look at the detailed x12arima output. The results of the "trading day" regression are fairly close to the top of the output. Having looked at this he may decide to 1) proceed without trading day adjustment, or 2) use a trading day adjustment, or 3) investigate further using x12arima directly. Best regards John On 9 October 2010 01:09, Allin Cottrell <cottrell(a)wfu.edu&gt; wrote: -- John C Frain Economics Department Trinity College Dublin Dublin 2 Ireland www.tcd.ie/Economics/staff/frainj/home.html mailto:frainj@tcd.ie mailto:frainj@gmail.com _______________________________________________ Gretl-users mailing list Gretl-users(a)lists.wfu.edu http://lists.wfu.edu/mailman/listinfo/gretl-users