Timeline for Independence of user actions on a social media platform
Current License: CC BY-SA 4.0
5 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Aug 15 at 22:07 | comment | added | jginestet | Say user1 has 1 month worth of data, and user2 has 10 months; you should divive the # of events (R oand S) by 1 for user 1, and by 10 for user 2. So the unit is "events per user-month" for all users. Now you can pick month, or week, or day or year... as your time basis. Or you could multiply the events for user1 by 12, and for user2 by 1.2 (to use 12 months, 1 year, as the time basis). Otherwise the weight of some users (long time users) will dominate. Note that you do not need to do this for regression; but you would need to do it for computing means/std, compare different groups, etc. | |
| Aug 15 at 22:00 | history | edited | jginestet | CC BY-SA 4.0 |
edited typos in body
|
| Aug 15 at 20:57 | comment | added | Amey Joshi | Thanks this answer was super helpful. I was thinking about it independently and was veering towards the same direction. There is one point, I didn't fully understand in your answer. What does "the users should be inversely weighted by their duration of observations" mean? | |
| Aug 15 at 20:55 | vote | accept | Amey Joshi | ||
| Aug 15 at 20:12 | history | answered | jginestet | CC BY-SA 4.0 |