I have a data that looks as follow . Post1 = 0 is collapsed mean of date -45 to -1 for the event date, post = 1 is collapsed mean of date 0 or the event date and post = 2 is collapsed mean of 1 to 45 after the event date. Surp is a surprise that is measured by actual - estimate / actual. So I have the form (numbers are arbitrary)

county_F event SURP POST1 y EA POST
1012 5/12 50 0 90 0 0
1012 5/12 50 1 95 1 0
1012 5/12 50 2 100 0 1
1012 8/12 45 0 80 0 0
1012 8/12 45 1 85 1 0
1012 8/12 45 2 78 0 1

My regression of interest is
y_{county, rdq} = surp + EA + POST + SURP#EA + SURP#POST.
to see where positive surprise effects my y. It felt like running the data above would create some trouble as all I am doing here is repeating the same data 3 times.

Now, I adjusted the data in the following form because the previous data was repeating exactly the same data 3 times (just different y) so I now have

county_F event SURP y0 y1 y 2
1012 5/12 40 90 95 100
1012 8/12 45 80 85 78
1012 11/12 47 x y z

Where y0 = pre event day y,y1 = on event day and y2 = post event day y

I have been thinking a while to figure out how to run the event study regression of this form. I tried regressing spending 0 , 1, 2 as the dependent variable but it does not seem like a event study anymore...and I don't think this is the ideal way to run this!

What are your thoughts? How can I use my new data to create an event study. Or would running the data on my old data make more sense?