I have two time-series of 2 different products x & y. They belong to the same main category, e.g.: Iphone 6 and Iphone 5. Maybe one product is the predecessor of the other product.

I tested for correlation between the sales of the 2 products and got r=.67. Correlation is calculated on the number of sales of the 2 products for the same month. Does this mean that there is a causal relationsship between the sales of the two products? Are both sales pattern the same? In this case I might be able to forecast one time series with the historical data of the other one.

  • $\begingroup$ Probably not, correlation does not imply causation most of the time. The reason for stock X and stock Y rise together might not be that rise of X cause rise of Y or the other way around. It might be that rise of Z causes the rise of both X and Y. Rise of X might be a good indicator (in a short time) that Y is going to rise soon. But that doesn't mean X causes Y. $\endgroup$ – hamster on wheels Aug 4 '17 at 21:43
  • $\begingroup$ By the way, why it takes a year for cross-validate to have an answer for this ... $\endgroup$ – hamster on wheels Aug 4 '17 at 21:44

You informed us that these are two extremely different products. If there is no theoretical reason at all for sales to be correlated and you still find a strong correlation, it is likely that there is a spurious relationship. I would like to remind you that this is a bivariate analysis, and the interpretation of the relationship between sales would remain limited unless you extend it to multivariate framework (by including other predictors, controls etc.).

After edit:

I think the argument still holds. Without accounting for other factors (with additional assumptions), looking only to correlation between sale numbers (for a month) to make predictions about future would be misleading. Maybe reading more on forecasting will be a better strategy, this website might be helpful.

  • $\begingroup$ They are 2 different products; but they are under the same main category.For eg: Iphone 6 and 6s. $\endgroup$ – kuppi Sep 29 '16 at 4:02
  • $\begingroup$ @kuppi So, don't you think that there might be a third factor which explains the correlation between the sales of these two products? And now you give the example of two phones, are these extremely different products? And if so, what is the criteria? $\endgroup$ – T.E.G. Sep 29 '16 at 4:04
  • $\begingroup$ They are 2 different products; but they are under the same main category.For eg: Iphone 6 and 5. They are 2 different products.they can be predecessor of the other product too(might or might not be).my aim is to find if their sales patterns are same; so that i can forecast the sales of 1 with the historical data of the other $\endgroup$ – kuppi Sep 29 '16 at 4:08
  • $\begingroup$ @kuppi It might be better to incorporate this new information (and other relevant ones) to the question, so others will have a better idea about your problem. $\endgroup$ – T.E.G. Sep 29 '16 at 4:12
  • $\begingroup$ I wanted to know if i can rely on the correlation; to say they are similar products or successor/predecessor; so that i can use one products data to predict the other.I do not have any other factors affecting it. $\endgroup$ – kuppi Sep 29 '16 at 4:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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