Came over from hd_prophet and find this fascinating! Really, really interesting. And hope you won't mind my butting in too terribly, but I do have question, mostly about sampling bias but also about patterns in fandom. Not hypotheses or even suppositions, but questions, possibly of the insoluble variety, which can safely be taken rhetorical musings-out-loud. (Though I am genuinely interested in any and all thoughts!)(And also, these definitely are not anticipating or requesting further data analysis! I'm still amply excited about the one set of graphs.)
- What is the influence of sampling from rec lists? You, capitu, and gracerene are all very well read, and I would wager more broadly read than many, and all read both top!Harry and top!Draco, so - as you say - I don't expect that your own preferences, as such, influences the data pool. But there are many fic writers out there who don't get recced (or read, or kudo'd - there's a tumblr post out there that's escaping me atm that shows that either a plurality or majority of fics have <5 kudos and very few hits). Is it safe to assume that what ends up on rec lists is reflective of what doesn't? - Relatedly, is the sample as large as it would seem to be? Your spreadsheet of capitu's rec list shows 1293 stories, but those are from only ("only") 444 unique authors, who may (or may not) have patterns in their writing that are less statistically significant than those of the very prolific writers you mention, but may still have some impact, especially if there's any sort of effect in the aggregate. - Is there a bias in which older fics were preserved and/or are still read? Are people who are reading and reccing now able to capture the zeitgeist of the community as it was then, since it included many works that are no longer readily accessible? - Would the results be different if based on a ratio instead of a percentage? As in, comparing the number of top!Harry to number of bottom!Harry stories, and number of top!Draco to bottom!Draco stories? A quick search of AO3 tags (which are not a good sample - tiny, inconsistently applied, not sorted by year - but is what's available quickly rn) shows that the ratio of top!harry to bottom harry is 1:1.2, but the ratio of top!draco to bottom!draco is 1:2. - Because there's potential for unknown sources of bias, random sampling is generally the safest bet, but is that meaningfully possible in a community that's spread over multiple platforms? - Is the shift over time reflective of more top!draco, or of more non-penetrative sexual activity? Is Draco topping more, or is topping increasingly irrelevant b/c of a shift towards frottage, wanking, oral, etc.?
no subject
- What is the influence of sampling from rec lists? You, capitu, and gracerene are all very well read, and I would wager more broadly read than many, and all read both top!Harry and top!Draco, so - as you say - I don't expect that your own preferences, as such, influences the data pool. But there are many fic writers out there who don't get recced (or read, or kudo'd - there's a tumblr post out there that's escaping me atm that shows that either a plurality or majority of fics have <5 kudos and very few hits). Is it safe to assume that what ends up on rec lists is reflective of what doesn't?
- Relatedly, is the sample as large as it would seem to be? Your spreadsheet of capitu's rec list shows 1293 stories, but those are from only ("only") 444 unique authors, who may (or may not) have patterns in their writing that are less statistically significant than those of the very prolific writers you mention, but may still have some impact, especially if there's any sort of effect in the aggregate.
- Is there a bias in which older fics were preserved and/or are still read? Are people who are reading and reccing now able to capture the zeitgeist of the community as it was then, since it included many works that are no longer readily accessible?
- Would the results be different if based on a ratio instead of a percentage? As in, comparing the number of top!Harry to number of bottom!Harry stories, and number of top!Draco to bottom!Draco stories? A quick search of AO3 tags (which are not a good sample - tiny, inconsistently applied, not sorted by year - but is what's available quickly rn) shows that the ratio of top!harry to bottom harry is 1:1.2, but the ratio of top!draco to bottom!draco is 1:2.
- Because there's potential for unknown sources of bias, random sampling is generally the safest bet, but is that meaningfully possible in a community that's spread over multiple platforms?
- Is the shift over time reflective of more top!draco, or of more non-penetrative sexual activity? Is Draco topping more, or is topping increasingly irrelevant b/c of a shift towards frottage, wanking, oral, etc.?