I am working on analysing advertising data for a digital advertising agency using Gretl.
The context of my current problem – I am analysing several campaigns over several months so this is panel data to find the impact of different days of the week on advertising conversion (users seeing an ad on the internet then clicking on that ad to then make a purchase from the website). I would like to analyse the panel using time fixed-effects models to see how the impact of Monday’s advertising differs to that of Tuesday’s.
My problem is that – I have several Monday, …, Sunday observations for each campaign as the advertising campaigns run over several months as shown below:
Date |
Day of Week |
Y |
X1 |
X2 |
X3 |
|
Campaign 1 |
02/01/2012 |
Monday |
. |
. |
. |
. |
Campaign 2 |
02/01/2012 |
Monday |
. |
. |
. |
. |
Campaign 3 |
02/01/2012 |
Monday |
. |
. |
. |
. |
Campaign 1 |
03/01/2012 |
Tuesday |
. |
. |
. |
. |
Campaign 2 |
03/01/2012 |
Tuesday |
. |
. |
. |
. |
Campaign 3 |
03/01/2012 |
Tuesday |
. |
. |
. |
. |
Campaign 1 |
04/01/2012 |
Wednesday |
. |
. |
. |
. |
Campaign 2 |
04/01/2012 |
Wednesday |
. |
. |
. |
. |
Campaign 3 |
04/01/2012 |
Wednesday |
. |
. |
. |
. |
Campaign 1 |
05/01/2012 |
Thursday |
. |
. |
. |
. |
Campaign 2 |
05/01/2012 |
Thursday |
. |
. |
. |
. |
Campaign 3 |
05/01/2012 |
Thursday |
. |
. |
. |
. |
Campaign 1 |
06/01/2012 |
Friday |
. |
. |
. |
. |
Campaign 2 |
06/01/2012 |
Friday |
. |
. |
. |
. |
Campaign 3 |
06/01/2012 |
Friday |
. |
. |
. |
. |
Campaign 1 |
07/01/2012 |
Saturday |
. |
. |
. |
. |
Campaign 2 |
07/01/2012 |
Saturday |
. |
. |
. |
. |
Campaign 3 |
07/01/2012 |
Saturday |
. |
. |
. |
. |
Campaign 1 |
01/01/2012 |
Sunday |
. |
. |
. |
. |
Campaign 2 |
01/01/2012 |
Sunday |
. |
. |
. |
. |
Campaign 3 |
01/01/2012 |
Sunday |
. |
. |
. |
. |
Campaign 1 |
02/01/2012 |
Monday |
. |
. |
. |
. |
Campaign 2 |
02/01/2012 |
Monday |
. |
. |
. |
. |
Campaign 3 |
02/01/2012 |
Monday |
. |
. |
. |
. |
Campaign 1 |
03/01/2012 |
Tuesday |
. |
. |
. |
. |
Campaign 2 |
03/01/2012 |
Tuesday |
. |
. |
. |
. |
Campaign 3 |
03/01/2012 |
Tuesday |
. |
. |
. |
. |
Campaign 1 |
04/01/2012 |
Wednesday |
. |
. |
. |
. |
Campaign 2 |
04/01/2012 |
Wednesday |
. |
. |
. |
. |
Campaign 3 |
04/01/2012 |
Wednesday |
. |
. |
. |
. |
Campaign 1 |
04/01/2012 |
Thursday |
. |
. |
. |
. |
Campaign 2 |
04/01/2012 |
Thursday |
. |
. |
. |
. |
Campaign 3 |
05/01/2012 |
Thursday |
. |
. |
. |
. |
Campaign 1 |
05/01/2012 |
Friday |
. |
. |
. |
. |
Campaign 2 |
05/01/2012 |
Friday |
. |
. |
. |
. |
Campaign 3 |
06/01/2012 |
Friday |
. |
. |
. |
. |
Campaign 1 |
07/01/2012 |
Saturday |
. |
. |
. |
. |
Campaign 2 |
07/01/2012 |
Saturday |
. |
. |
. |
. |
Campaign 3 |
07/01/2012 |
Saturday |
. |
. |
. |
. |
Campaign 1 |
08/01/2012 |
Sunday |
. |
. |
. |
. |
Campaign 2 |
08/01/2012 |
Sunday |
. |
. |
. |
. |
Campaign 3 |
08/01/2012 |
Sunday |
. |
. |
. |
. |
When I put the dataset into Gretl and run a panel data regression, it does not give me 6 (7-1) time dummy variables that are Tuesday, …, Sunday but it gives it to me as 89 (90-1) time dummies that are 02/01/2012, …, 30/03/2012. I don’t know whether this is the case in Stata.
I really would like to know a way to get time dummies for days of the week as opposed to the date. I want to find a way of getting these days of the week dummies, without having to average out all the say Mondays’ Y, X1, …, Xn over the several months to get one average Monday’s Y, X1, …, Xn. Could a Monday index variable for all Y, X1, …, Xn for each campaign’s Mondays which takes into account all the Mondays’ be created?