reg or autoreg or arima

[podarum]

Ok, my scenario is like this.  I have prices of certain (5 types) cars that go as back as 15 years by month, and by zipcodes (30 of them).  I also have some other variables such as household income and unemployment %, which I want to attempt to use in forecasting the changes in auto prices in the next 12 months.  Arima woks, I know, but I'd have to do a model for each auto type (5) per each zipcode (30) = 150 models.  Can I use Regression and just get one model and have the same variables in the model and would it work?.. how about AUTOREG, since I don't know much about it?  thanks

[dekelas]

Hi,

I believe what you are looking for is The arimax method (which is the same as arima but with independent variables).

I would recommend inquiring each series separately so you can look at the structure of each series (unless you are looking for the average price of all car models.).

BTW: why not using the difference between each time period ? (T2/T1-1)

I SUGGEST YOU GOOGLE ABOUT ARIMAX. 

about AUTOREG - it is a simplier form of arima (AUTOREGRESSION-INTEGRATED-MOVING_AVERAGE)

hope this helps.

keep posting

[podarum]

Thank you dekelas, very helpful info.

When you say  difference between each time period, do you mean as the dependent variable?, and for both ARIMA and AUTOREG ?

Also, do you think just a Multiple Regression model would work fine with this data, and just use time as an independent variable..??

[dekelas · Accepted solution]

Hi,

I'm getting the feeling you don't fully understand the meaning of Time-Series analysis.

here is a nice start :

http://www.abs.gov.au/websitedbs/d3310114.nsf/4a256353001af3ed4b2562bb00121564/b81ecff00cd36415ca256...

and a nice article about arima models:

http://www.duke.edu/~rnau/411arim.htm

now to answer your question: I do mean the dependent variable (which is aso the independent variable in AR models). and you can use it for both autoreg and arima.

take care