How to detect a time deferred cause and effect

Hey guys,

I am looking at a fairly easy problem (I think). I have to columns of data, marketing spent and revenue. I have the feeling that when marketing spent goes up revenue increases about 5 to 10 days later. The graphs support my view I feel.

What statistical analysis could I use to a) affirm my believe, b) find event the magnitude of the effect and c) have an average number of days when the effect kicks in.

Would love to hear your thoughts on this!




Less is more. Stay pure. Stay poor.
Hmm, good question. I haven't done anything like this. But plotting this as a time series seems appropriate. Do you have multiple times where spending occurred, so repeated occurrences of this phenomenon? If so how many? Also, do you have a possible control group for the same time period that didn't have more marketing to show that it doesn't go up without the marketing?
The spending occurs every day in different intensity. The phenomenon should be (in theory) ongoing. Revenue is believed to increase by x% 3-7 days after the spending increasesd about y %.

There is not control group unfortunately.


Less is more. Stay pure. Stay poor.
I wonder if you may be able to tap into infectious disease literature. You get innoculated, have an incubation period, then manifest symptoms.


Fortran must die
You can run a time series model where you use an X ten days before the Y. Time series regression is not, however, for the faint of heart. Be sure to look at the issue of autocorrelation before you run such a test.