I am looking for if there is any statistical method(s) to view if there exists any relationship between counts. I have a large astronomical type of dataset but for the sake of breaking it down for an example I have modified the data to the following:

Event1         Event2      Event1_Count_Occured      Event2_Count_Occured         Count_Both_Events_Occured_Together       Total_Count_Of_All_Possbile

EventA         EventB             30                         40                                    19                                          60
EventA         EventC             30                         10                                    19                                          60
EventA         EventD             30                         60                                    19                                          60
EventB         EventA             40                         30                                    19                                          60
EventB         EventC             40                         10                                    19                                          60
EventB         EventD             40                         60                                    19                                          60
EventC         EventA             10                         30                                    19                                          60
EventC         EventB             10                         40                                    19                                          60
EventC         EventD             10                         60                                    19                                          60
EventD         EventA             60                         30                                    19                                          60
EventD         EventB             60                         40                                    19                                          60
EventD         EventC             60                         10                                    19                                          60

Essentially throughout a 60 day period (Total_Count_Of_All_Possbile) Event 1 occurred X amount of times (Event1_Count_Occured) and Event 2 occurred X number of times (Event2_Count_Occured).

In that 60 day period, some events occur on the same day as other events. To see how many events occur on the same day together that is denoted with the Count_Both_Events_Occured_Together field.

In my real dataset, there is more than just 4 events (Event A,B,C and D). There are probably ~8,000,000 unique events. I am trying to have some statistic(s) to see if there is a possibility based on the counts and the number of times they overlap to see if there is some possibility that some sort of relationship occurs. For example, if Event 1 occurred 2 times and Event 2 occurred 2 times in a 60 day period and they both occurred on the same two days I would say there is a possibility there is a very strong possibility of a relationship there since they both only occurred 2 out of the possible 60 days and they happened to occur on the same two dates. I would like some sort of way to measure this, but can't find any methods to accomplish this. I can't do any sort of standard correlation test since I just have the counts.

  • $\begingroup$ How is it possible that Even 1 + Event 2 > Total of All Possible Accounts as shown in your data? $\endgroup$ – StatsStudent Jan 30 at 23:04
  • $\begingroup$ @StatsStudent It is not the sum of both events. It is only the amount of time each event occurred. So for example in a 60 day period, Event A occurred 30 times and Event B occurred 40 times. So multiple events could happen on the same dates. The 'Count_Both_Events_Occured_Together' field counts how many times this occured. Total_Count_Of_All_Possbile refers to just the 60 day period. All rows will have the same value for this field. $\endgroup$ – user2924127 Jan 31 at 0:00
  • $\begingroup$ How large is this dataset by the way? With 8 million possibilities of Events, I'm guessing this dataset is massive? $\endgroup$ – StatsStudent Jan 31 at 0:03
  • $\begingroup$ It is. We are combining all events with each other so roughly 8000000 * 8000000 rows $\endgroup$ – user2924127 Jan 31 at 0:22

Your Answer

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

Browse other questions tagged or ask your own question.