I have two time series:

  1. numbers of subscriptions per hours for 10 days, for a website.

  2. the time where TV ads, for a specific TV channel, have been displayed e.g. 10:25AM on Monday 12 of August, TV AD1 displayed on CNBC as an example. The issue with that time series is there is only the time and the date but nothing else.

I would like to do a cross correlation to see the impact, or the lack of impact, of the TV ads on the numbers of subscriptions.

But with only the time of the adverts, is this cross correlation doable? If yes, how?

I'm not asking for a specific solution designed by any of you, rather website links where I can find some ideas or any hints as I'm stuck at the moment.


1 Answer 1


The actual date /time/channel is an observation/transaction. A time series is a bucketing of transactions. For each type of advert,I would take the time of the advert and bucket them into hours to create a number of possibly "causal" time series. The impact of an advert may depend on the hour of the day or the day of the week or whether or not it is on a holiday or even “nearly” on a holiday.Since you only have 10 days it is not feasible to try and compute daily/holiday effects. The cross-correlation between each of these discrete advert time series can be computed (descriptive statistic) but shouldn’t be as your predictor varaibles are discrete counts and your subsription data may be autocorrelated. I would create 23 predictor series reflecting hour of the day and include these as well as the advert time series computed above into an ARMAX model.Care should be taken to identify and deal with any subscription readings that reflected either Pulses, Level Shifts, Time Trends or Seasonal Pulses as these would be assignable to omitted variables that you had not controlled for. Hope this helps.

  • $\begingroup$ Hi Dave, It is indeed helpful. I will build these 23 predictive series based on the time the adverts are shown. Thank you. $\endgroup$
    – Andy K
    Commented Aug 12, 2013 at 10:41
  • $\begingroup$ The 23 predictive series have nothing to do with the time that the adverts are shown They simply are 23 0/1 variables reflecting the hour of the day. These 23 will be used in addition to the time series variables that I described to estimate the impact of the type of advert in addition to the ARIMA structure. $\endgroup$
    – IrishStat
    Commented Aug 12, 2013 at 12:47
  • $\begingroup$ I add the dummy variables in addition to my time series. I think it is what you mean... $\endgroup$
    – Andy K
    Commented Aug 12, 2013 at 13:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.