>>12705499Sure, it's getting late to start downloading a whole dump and editing code but I'll check tomorrow.
I'm more a Data Scientist than a programmer but I imagine if I can use R a python script can't be that hard to understand.
However the point was that, as you love data so much, you should learn more about analysis itself. Contextual understanding, skepticism of existing assumptions and so on. I'm not saying your assumption is right or not, but you made clear you had one before even running those scripts, and it's missing a ton of contextual data. Making an analysis to prove an opinion will always harm the quality of the result.
Your initial assumption seems to be that people mostly use discord to reply to beatani/get her attention and such, similar to what people complained about for twitter. In short, the effect of poster X being vtuber beatani.
It's pretty obvious and understandable that people are more likely to reply to the person the whole thing is about, in the first place. However, to properly measure only that effect, you have to get ride of other parasite factors, and there are many : beatani tends to post to start topics related to her channel (ex: minecraft events) which makes a channel goes from dead to active, beatani posting style is A LOT of short messages in serie, which boost the activity by a ton, beatani talks about stuff that interest more people, which is logic considered users are fans of her content, etc.
Ideally each of those factors that lead to an increase in activity should be measured separately if you want to stricly show the effect "being able to reply to beatani" have on the activity.
Yes, showing something so precise is indeed a pain in the ass, often one result is due to many different factors and that's why it's a full time job for more serious topics.
