# regression time series

#### masarimk

##### Member
Dear Members,

I need to do time series analysis. I have retail sales data of a retail company since January 2006(every month). I need to forecast 2012. But I have other variables that effect the sales such as number of stores, supply ...It seems time is not only the variable that has impacts on forecast. Can anybody suggest me to any models? I was thinking multiple linear regresision but how do I hande time and other variables? How do I model time series and linear regression at the same time? I appreciate with any assistance.
Thank you.

#### noetsi

##### No cake for spunky
According to a proffessor of mine, multiattribute ARIMA is the "simplest" way to determine the relative effect of multiple time time series on another (if that is what you are trying to do). Having said that (this is my own opinion from what I have seen in discussions of ARIMA) even the simplest ARIMA is complex and involves judgements on the data. Nor have I found much on this form of ARIMA - it's just a suggestion of where you might look.

You might look at the following article on ARIMA to see if it meets your needs (this is for one time series on one time series). If so, start on page 380 or so that discusses the methods - before that they are talking about the specific topic that has nothing to do with your interest.

Police Shooting and Criminal Homicide: TheTemporal Relationship Robert H. Langworthy'

it's in Journal of Quantitative Criminology, Vol. 2, No. 4, 1986

This might also be of some help (as an introduction to time series generally based on my own painful process to date)

Time series
A 9.2 tutorial
R
http://decisionstats.wordpress.com/2010/10/18/using-r-for-time-series-in-sas/
SAS/ETS Software: Time Series Forecasting system
Fidel Time Series chapter
http://wps.ablongman.com/wps/media/objects/2829/2897573/ch18.pdf
http://www.otago.ac.nz/sas/ets/chap2/index.htm
A book on time series for SAS by two well known statisticians (heavy on the math).
http://www.scribd.com/doc/46954465/SAS-for-Forecasting-Time-Series-Second-Edition
“The Time series procedure creates the required Jurisdiction-level totals for each month. It’s really aimed at aggregating transactional data within intervals, but it will also work to aggregate within BY groups. Easy way to code this in SAS Enterprise Guide: select the data set, go to the pull-down menus and select Analyze -> Time Series -> Create Time Series Data. This will bring up a wizard which asks you to specify the variable roles, and how the time series variables will be treated. Run this, then open the ‘Last Submitted Code’ project node to get pre-written code that can be modified to suit.”
How to create time series data.
http://excelicious.wordpress.com/2010/03/25/time-series-analysis-in-sas/
An excellent review of time series regressed against another time series.
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2596667/

#### jpkelley

##### TS Contributor
You can also try Generalized Additive Models.