arima multiple periods

[masarimk]

Dear All,

I have t-shirt sales data from week 14 to 40 weeks for 2011 and 2012, thus I have 50 data weeks(2012 is not completed yet). I defined my time series data with frequency 27 . But data hierarchy is years , months, weeks.

Question 1) In arima I defined period 27 for the week. Do I need to define dummy variables for the year 2011 and 2012 and months? Will Arima catch trend between 2011 and 2012? And also note that I can not remove trend and seasonality by subtracting 1 and 27 from data since I don't have sufficient number of data points.

Question 2)
My R codes:
xregmodel has dumys for year and months.
auto.arima(dftrain,xreg=xregmodel,freq=27,lambda=0) but this does not give me any p values and r square. Is there a way to report them?

[masarimk]

I was not able to get an answer for this question above but maybe I need to rephrase it;
I have weekly sales data but also year and months also show some correlation. One example is every July T-shirt sales are always similar beacuse as the weather gets hot, the sales increases. So do I need to use arima with multiple period? If I only use Arima with period 52 for the weeks and dummys for months and years? or is it too much too use maybe use dummys for years?
Thank you

[derksheng]

This might interest you:

http://papers.ssrn.com/sol3/papers.c...act_id=1989593