The Harmon Foods case has also Irregular expenditures. That's why I mentioned it. I suppose you mean a Koyck model? Why not a distributed lag model ? You will well see if there is a geometric decay. In the case of Harmon Foods the product was cereals for breakfast ("corn flakes") so when households had bought it one month they were not inclined to buy next month, producing a negative coefficient for lag 1. With Koyck you will miss that. Be ready to include ARMA errors. Moreover in the case they had two kinds of expenditures: those directed to consumers and those directed to distributors. A group of my students had shown fitted values using each of them, or only time variables, or everything together. I have kept these four plots in my course to show the power of regression
Guy Melard Thanks for you time & sugession! Sir, if I use simple "Distributed Lag Model" instead of "Koyck Model" than I think there might come the problem Multicollinearity between lag variables & also how can I know how much lags i should take...The harmon Foods example seems interesting but right now, in my data on the dependent side (i.e.sales data) I have value for all data point say for 50 weeks but on Independent side (i.e. advertisement) I have have only data for 15 -20 weeks even it is also not regular (i.e. ads for 4-5 weeks then gap for 6-8 weeks then again 5-6 week like this )..Please suggest...I am referring the book Damodar N. Gujrati Basic Econometric, is it be helpful?
Thanks & Regards