Date: Tue, 3 Feb 2009 08:24:15 +0000
Reply-To: franz_cl2003@yahoo.fr
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: Franz <franz_cl2003@YAHOO.FR>
Subject: Re: Learning (please help)
In-Reply-To: <1115a2b00902021722v453df8e1i6c4a39a03e472a@mail.gmail.com>
Content-Type: text/plain; charset=us-ascii
Hi Wensui,
The proc kde step on my original data (having only 9 observations) has generated an output nb1_kde (with 3600 obs!!). That's the reason why I put nlag = 40 in the following proc arima.
Since I am still learning about the whole process I would really appreciate any guidance/suggestion.
Thanks & Kind regards,
Franz
--- On Tue, 2/3/09, Wensui Liu <liuwensui@gmail.com> wrote:
> From: Wensui Liu <liuwensui@gmail.com>
> Subject: Re: Learning (please help)
> To: franz_cl2003@yahoo.fr
> Cc: SAS-L@listserv.uga.edu
> Date: Tuesday, February 3, 2009, 2:22 AM
> Hi, Bro,
>
> There is onething I don'n understand in your code.
>
> for 9 data points, how could you put nlag = 40 in your proc
> arima?
>
> On Mon, Feb 2, 2009 at 7:01 PM, Franz
> <franz_cl2003@yahoo.fr> wrote:
>
> > Dear All,
> >
> > 1.-To smooth my data I have used proc kde this way:
> (Any other suggestion
> > is welcome)
> >
> > proc kde data=tennis gridl=1 gridu=20 method=snr
> out=nb1_kde;
> > var nb1 t;
> > run;
> >
> > 2.- Then I have partitioned my output nb1_kde into
> various groups and made
> > sure that the plot of the MEAN against the SD is
> constant accross the
> > groups.
> >
> > 3. The following code then outputs the resulting
> autocorrelation function
> > and plot.
> >
> >
> > proc arima data=nb1_kde;
> > identify var=nb1 outcov=auto nlag=40;
> > run;
> > quit;
> >
> > proc gplot data=auto;
> > symbol1 i=needle v=none width=5;
> > title1 "Autocorrelation plot";
> > plot corr*lag;
> > run;
> > quit;
> >
> > I would be nice to have some advices.
> > Thanks & Kind regards,
> > Franz
> >
> >
> > --- On Mon, 2/2/09, Franz
> <franz_cl2003@yahoo.fr> wrote:
> >
> > > From: Franz <franz_cl2003@yahoo.fr>
> > > Subject: Re: Learning (please help)
> > > To: SAS-L@listserv.uga.edu
> > > Cc: "Wensui Liu"
> <liuwensui@gmail.com>
> > > Date: Monday, February 2, 2009, 11:08 AM
> > > Dear all,
> > >
> > > Well, let's say that the goal is to forecast
> membership
> > > by 2011.
> > > I have been actively trying to
> "normalize" my
> > > data (see code bellow).
> > > Would you experienced guys think I am on the
> right path?
> > > Any advice is well come.
> > >
> > > data tennis;
> > > input year nb;
> > > t = _n_;
> > > nb1 = log(nb);
> > > cards;
> > > 2000 8
> > > 2001 11
> > > 2002 14
> > > 2003 25
> > > 2004 30
> > > 2005 45
> > > 2006 42
> > > 2007 130
> > > 2008 163
> > > ;
> > >
> > > * series plot *;
> > >
> > > proc gplot data=tennis;
> > > plot nb*t;
> > > symbol i=join;
> > > run;
> > > quit;
> > >
> > > %macro test (z=, dat=);
> > >
> > > * Histogram of the Series *;
> > >
> > > proc univariate data = &dat noprint;
> > > histogram &z / normal(noprint);
> > > inset
> > > n = "N"(5.0)
> > > mean = "Mean"(5.0)
> > > median="Median"(5.0)
> > > std = "Std Dev" (5.0)
> > > SKEWNESS="SKWNESS"(3.1)
> > > KURTOSIS="KURTOSIS"(3.1)/
> > > pos=ne
> > > height = 1
> > > header = 'Summary Statistics';
> > > axis1 label=(a=90 r=0);
> > > run;
> > >
> > > * Density *;
> > >
> > > proc capability data= &dat;
> > > var &z;
> > > HISTOGRAM / kernel( k=NORMAL c=MISE
> > > color=BLUE l=1)cfill=GRAY;
> > > run;
> > > %mend test;
> > >
> > > %test (z=nb, dat=tennis);
> > > %test (z=nb1, dat=tennis);
> > >
> > > * Smmothing (nb)*;
> > >
> > > proc kde data=tennis gridl=1 gridu=20 method=srot
> > > out=nb_kde;
> > > var nb;
> > > run;
> > >
> > > %test (z=nb, dat=nb_kde);
> > >
> > > Thank you very much & Kind regards,
> > > Franz
> > >
> > >
> > > --- On Mon, 2/2/09, Wensui Liu
> <liuwensui@gmail.com>
> > > wrote:
> > >
> > > > From: Wensui Liu <liuwensui@gmail.com>
> > > > Subject: Re: Learning (please help)
> > > > To: franz_cl2003@yahoo.fr
> > > > Cc: SAS-L@listserv.uga.edu
> > > > Date: Monday, February 2, 2009, 2:46 AM
> > > > if you have only 9 data points, then
> don't bother
> > > to
> > > > waste your time on
> > > > arima or other fancy models or software. if
> i were
> > > you, i
> > > > will just do a
> > > > simple univariate smoothing.
> > > >
> > > > On Sat, Jan 31, 2009 at 6:07 AM, Franz
> > > > <franz_cl2003@yahoo.fr> wrote:
> > > >
> > > > > Dear All,
> > > > >
> > > > > Below are membership data from a Tennis
> club...
> > > > > I am trying to model the Time Series in
> order to
> > > make
> > > > some predictions.
> > > > > Due to the huge increase of membership
> number
> > > from
> > > > 2006 to 2007, a plot
> > > > > of Residual vs Fitted Value reveals
> some
> > > outliers.
> > > > >
> > > > > 2000 8
> > > > > 2001 11
> > > > > 2002 14
> > > > > 2003 25
> > > > > 2004 30
> > > > > 2005 45
> > > > > 2006 42
> > > > > 2007 130
> > > > > 2008 163
> > > > >
> > > > > How do I have to deal with those
> influential
> > > points in
> > > > this specific case.
> > > > > I would be great to have some
> explanation/code
> > > too,
> > > > since I am new to the
> > > > > topic.
> > > > >
> > > > > I just know that some adjustments have
> to be
> > > done. I
> > > > have been reading
> > > > > about procedures (ARIMA, EXPAND, MIXED,
> UCM ...)
> > > and
> > > > don't really know how
> > > > > to proceed.
> > > > >
> > > > > Many Thanks & Kind regards,
> > > > > Franz
> > > > >
> > > >
> > > >
> > > >
> > > > --
> > > > ===============================
> > > > WenSui Liu
> > > > Acquisition Risk, Chase
> > > > Blog : statcompute.spaces.live.com
> > > >
> > > > I can calculate the motion of heavenly
> bodies, but not
> > > the
> > > > madness of
> > > > people."
> > > > -- Isaac Newton
> > > > ===============================
> >
>
>
>
> --
> ===============================
> WenSui Liu
> Acquisition Risk, Chase
> Blog : statcompute.spaces.live.com
>
> I can calculate the motion of heavenly bodies, but not the
> madness of
> people."
> -- Isaac Newton
> ===============================
| 
