Date:         Wed, 2 Mar 2005 22:27:49 +0100
Reply-To:     Grado <grado2005@gamespy.com>
Sender:       "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From:         Grado <grado2005@GAMESPY.COM>
Subject:      Re: Time series: do i use it in this case?

<express_myself_01@yahoo.com> ha scritto nel messaggio
news:1109796324.017564.281120@z14g2000cwz.googlegroups.com...
> Hi...
>
> I have a question, maybe someone can answer.  I have data which is
> population data for 10 successive year (number of people in a certain
> age group), which is used to predict company sales for those same
> years.
>

Hi,

I must say first of all that I'm not an expert in time series analysis and I
don't know a lot of forecast techniques, however I'm pretty sure that is
hard to find out a consistent model with only 10 periods time series.

I have a  little insight in ARIMA models, but they do not seem to fit your
problem cause they need long series (more than 70/80) to assess the
existence of AR or MA components.

Let me understand better what is your problem: I think you have set up a
model where company sales are a function of population, and now you wish to
predict the independent variable in order to have a forecast of future
sales. Is it correct?
If so, maybe you can try to look for a relationship between population and
some other socio-economic variables, and maybe you can find longer time
series for these variables; at this point you could some sort of time series
model (ARIMA, or model for seasonal-trend decomposition).

Bye