Species Popularity by Sex, Gender & Sexual Orientation

We have a new visualisation to share today, courtesy of the industrious and talented hooves of Ruxley (https://github.com/ruxley).

This is an interactive visualisation which lets you explore the popularity of the top furry species, and see how that popularity changes with biological sex, gender, and sexual orientation.

Exactly how many wolves are there? (Lots and lots.) Are foxes gay? (Not really.) Are horses more popular than zebras? (Duh.)

This data is taken from the Furry Survey. You can set a baseline dataset, and compare this to any other dataset. So you can directly compare men with women, heterosexuals with homosexuals, or any combination with the wider furry community. It is, I think you will agree, quite nifty.

There are a few interesting discoveries to be made, some surprising and others not-so-surprising. We will leave it to you to explore, although of course we at [a][s] are sure to reference this visualisation in future articles.

As a starting point, we suggest that you compare heterosexuals to homosexuals… and look at the otters.

Be aware that the data becomes less reliable if you select a small dataset. This is a normal outcome – just don’t take results from small datasets too seriously.

The more observant of you will notice some unusual results from the data. This is because many furries choose multiple species. For the purposes of this data, we count each named species once: so a fox-wolf hybrid counts as one fox and one wolf, and a furry who is usually a raccoon, has a tiger alt, and observes Dragon Friday will be counted three times.

Curiously, there are big variations in the number of species chosen by each of our groups. (This is why the data may behave oddly when you compare different groups.)

Women choose many more species (2.0 per furry by sex; 1.9 by gender) than men (1.5 by sex; 1.4 by gender). And heterosexual furries choose more species (1.7 per furry) than homosexual furries (1.4 per furry).

Why? No idea. But it sounds like an [adjective][species] type of question.

About JM

JM is a horse-of-all-trades who was introduced to furry in his native Australia by the excellent group known collectively as the Perthfurs. JM now helps run [adjective][species] from London, where he is most commonly spotted holding a pint and talking nonsense.

Before posting a comment, please read our Code of Conduct

19 thoughts on “Species Popularity by Sex, Gender & Sexual Orientation

  1. You may want to treat multiple-species responses as contributing only fractionally to each species (e.g., 0.5 wolf and 0.5 fox instead of 1 wolf and 1 fox for a fox+wolf response). A quick look at the male/female plot suggests that females overrepresent males in almost every species except foxes and huskies, but (reading your discussion) this just seems to reflect females choosing more species and therefore being multiply-counted. It would be easier to judge relative species popularity if the columns added to 100%.

    1. Hi Procyon, well observed, we noticed the same thing, which is what led us to investigate the male/female multiple species choice difference. Unfortunately with the current data we have we’re unable to distinguish between species choices that are part of the same respondent, but I agree that it would help make the graph more accurate.

  2. Procyon’s suggestion sounds worthwhile to me.

    I like that you can compare two different groups directly, but I wish I could also directly compare the three orientation that were listed.

    1. Hi keito, unfortunately with the data we have we’re unable to make the distinction between multiple species per respondent, as mentioned in the reply to Procyon above. I agree that comparing all three orientations at once would provide a better overview, but it would also make the graph larger and more visually complex.

      I have open-sourced the visualisation on GitHub, so if anyone would like to fork it and make such a modification then I’d love to see it!

    1. Hi Fluffy. Good question, and it’s something I should have covered in the article itself.

      When we ask for sexual orientation, we give seven options spanning the heterosexual-homosexual binary (from “completely heterosexual” through to “bisexual” through to “completely homosexual”), plus “asexual” and “pansexual”. This is too many categories for the visualization, partly because the data has less value when it is based on a small number of datapoints, partly because the visualization itself becomes unwieldy.

      We made the decision early to have three options each for sex, gender & orientation. To manage the orientation data, we lumped together responses into the three categories you see (including pansexuals in with the bisexuals, which at best is a very crude simplification). We discarded the asexual data.

      Asexuals make up about 5% of furry, which is significant but small from a data analysis point of view. We judged that there wasn’t much value in showing this due to the unreliability of small datasets. In hindsight, this was the wrong decision. We should have shown this data because (1) it’s interesting and (2) asexuals are unfairly ignored often enough, and we’re contributing to the problem.

      We got this one wrong. We don’t intend to make the same mistake in the future. We have more visualizations coming.

      1. I’m actually surprised to hear that the asexual furry population is 5% of the sample set overall. It’s nice to get that level of data, at least.

        How big is the original data set in the first place, and how was it obtained? It seems like there’s always going to be a lot of sampling bias and there will need to be error bars in order for any comparison to be really meaningful.

        And, thanks for understanding the issue of ongoing erasure.

        1. The data is taken from the Furry Survey, which is an online survey that’s been run by [a][s] founder Klisoura for the last decade or so. It receives several thousand responses each year. (2014 is coming soon!)

          There is bias, in that our sample is self-selecting. However we see consistency in our results regardless of how people have find the survey (via LJ, Twitter, or more recently ads on FA), and the results have been shared with and validated by the professional researchers at the IARP. They think that our survey data is a good and full reflection of the whole furry community, a conclusion based on comparison with their own data gathering at conventions and online, although of course as with any sampling bias there is no way to quantify this. Still, if the professionals think the data is of very high quality, I’m not going to argue the point.

          The use of error bars is another good question and one we struggle with. In general, because our audience is lay, we at [a][s] tend to err on the side of simplicity when presenting data. That’s what we have done here.

          That has shortcomings: if you fiddle with the visualization to show only sex “other” (i.e. intersex furs) you’ll see that the results are pretty meaningless because there are so few people who fit into that category. I decided to manage this with a note in the text rather than trying to include some measure of statistical significance in the visualization itself. Essentially we chose clarity over detail (we had long discussions over whether the visualization might be too complex as it is).

          Dr Courtney Plante (aka Nuka), who is a furry and writes for us here at [a][s] on occasion, has looked at the various trade-offs in the sort of research we do at [a][s] and he does at the IARP. It’s an enlightening read:

          http://www.adjectivespecies.com/2014/04/22/trade-offs-in-furry-research-adjectivespecies-vs-the-iarp/

          I’ll add that I have a science degree with a few publications and also am trained in the proper use of statistics. Nuka and I have been chatting at length in recent weeks about our combined datasets, and you’ll be seeing more from him (and, of course, me) on this topic in coming weeks and months.

          1. Sounds good! I’m probably at the “knows enough to be dangerous” level with statistics, myself, so I’ll trust your judgement.

            Although I’d mention that the way you talk about your degree does feel a bit off-putting, since there seems to be an implicit assumption that the person you’re talking to doesn’t also have an advanced degree in science – in my case, computer science, which I was close to completing a PhD in but I decided to quit academia with “only” a master’s when I got sick of the process. You might want to work on not sounding like an ivory-tower gatekeeper when engaging with people you don’t know.

          2. Incidentally, general-population estimates put asexuality at around 1%, which is why I find the stated 5% of the furry population to be so interesting. Which is another reason why having even that information as a note on the data would have been nice. I’m sure it was published elsewhere, but this was the first post I’d seen on this survey or its results, and there wasn’t a link to the greater survey in the article as far as I could see.

            Also, given that you DO allow selecting intersex as an option, and you point out how useless the visualization is there, why not also allow selecting asexual as an option and just provide a similar note as to its futility? I note that a gender of “other” shows 42 responses, or about 0.5%; if there truly are 5% responses categorized as asexual then that’s 10x as many.

            So, would you consider at least adding asexual as a dropdown choice on this one just for now?

        2. (Replying here because we’ve reached the maximum comment depth.)

          Thanks for the heads up on my language. My intent isn’t to pull rank or be off-putting, apologies for coming across that way. I’m a bit defensive because we get challenged about our use of stats regularly, and in fact that was the original motivation for that article from Nuka.

          Many people, when faced with a result in the data they don’t like, often choose to attack our use of numbers. (I know that’s not what you were doing.) And so we find ourselves having to defend our knowledge and, yes, qualifications.

          Thanks for still taking the time to respond in a friendly and open manner despite my language.

          1. Sure, it’s something we all need help with from time to time. :) I’ve certainly been guilty of it myself in the past.

            I’m just genuinely interested in the statistics, and being asexual I couldn’t help but feel a bit unrepresented and therefore erased by that decision.

        3. “So, would you consider at least adding asexual as a dropdown choice on this one just for now?”

          Unfortunately that data was lost when we originally extracted it, and formatted it in preparation for this visualization. So we can’t add it as an option without going back to square one. You can take a look at Ruxley’s Github repository to see what we’re up against.

          1. Ah, that’s unfortunate, but understandable. I know that data preprocessing often comes with hard decisions, and rerunning the preprocessing on a data set is often too annoying to be worth it for these sorts of changes. And of course, releasing the unprocessed data set opens up massive privacy issues.

            I’ll definitely be interested to see what other ways the data gets sliced in the future, though!

    1. Hi Tinker, thanks for the heads-up. This is an intermittent, ongoing problem – our host has been under concerted attack for some time. It’s nothing to do with [adjective][species]. I’ve passed on your message to the site owner (by sending a message to @adjspecies) – it normally gets sorted out quickly. Thanks.

Leave a Reply

Your email address will not be published. Required fields are marked *