Regression and seasonality

[statsgirl123]

I am running a multiple linear regression analysis to show how different demand factors affect demand of juice. My independent variables are price, temperature, competitor price, feature space and competitor feature space. However, I know there is an element of seasonality which is affecting the demand. I am looking at 2 years worth of weekly data, so feel that including dummy variables isn’t an option. I have seasonality indices which have been loosely worked out using an average of 4 years, removing sales caused by promotion and giving a percentage increase or decrease from how much sales vary on average from the base sale. How can I incorporate seasonality into my regression model?

 

[hlsmith]

So for clarification, you do not think your scenario would fall under time series?


Have you plotted the relationship to see if their appears to be a relationship you may need to control for?

 

[statsgirl123]

I wouldn't use a time series model here as I'm trying to measure to what extent external factors are influencing the demand - there just appears to be a general 'spike' in the demand at certain points during the year

 

[CB]

Maybe a trigonometric (e.g., sine/cosine) function of time would be useful here as a control variable? That'd allow you to control for some types of smoothly shifting seasonal demand.

 

[hlsmith]

I was also thinking of a spline term.