Glad you liked it! Although, since another commentor pointed this out, do be careful in taking the end result too literally. I avoided using numbers because I want the viewer to draw their own conclusion of the debate while keeping these things in mind. Though I do have to admit, my video is obviously biased against bears.
When I heard about the man vs bear question, I thought the point was that men are *perceived* as more dangerous than bears. If men really are less dangerous than bears, what can men do to change the false perception?
I know this Simpson's man he was Northern Irish & he made the backwards Simpsons significance index thing. that my Northern Irish exame board uses because well we can't use the english inverse version.
I cant believe this is how the general public gets exposed to statistics now lol Actually its not that bad, but i think its funny that some people genuinely started researching and thinking about statistics because of this
Yeah, the numbers are a bit outlandish because I didn't want people to focus on the numbers; I wanted people to focus on the concept. This video is a biased because it uses mostly anti-bear examples, and while I didn't mean for it to turn out that way, in my (likely biased) opinion, I thought these examples were more agreeable and easier to see the statistical error in. Some other examples could've included: Bears encountered at the outskirts of forests (i.e. near urban/foresty borders) might (?) be more aggressive due to a number of potential factors (e.g. desperation for food, being pushed out from their territory, actively seeking out humans, etc.) and NOT a good representation of the overall bear population. Since these bears are closer to human civilization, if these bears are indeed more aggressive, then we would have a Simpson's Paradox-like effect when combining this population of bears with other bear encounters. Bears encountered by people trained to track/deal with bears will result in fewer accidents than people randomly out in the woods, so including these encounters might unfairly nudge the numbers the other way (in favor of bears). Since trained professionals aren't representative of the average human population. There are a lot of social/biological factors that impact human-human interactions that could go either way (pro/anti male). These factors probably aren't relevant to bears (you're either food/not food) and if they are, we have no way of accurately checking (we don't speak bear). What constitues an encounter? Bears can smell very far and might determine that you're food that's not worth chasing after. Does walking past a man on a dimly lit street for 1.5 seconds count as a male encounter? What if that man was across the street? In my opinion, there is a huge gray area in terms of effective encounter range which make it hard to pin on precise and agreeable numbers. Points like these carry a lot more nuance, which I thought might distract from the statistical principles mentioned in the video. However, this led to my mistake of publishing a video that was anti-bear. When my intention was to shine light to the fact that saying a true statement like "you are times more likely to be attacked by a man than a bear" can potentially be misleading if you don't look at all the relevant statistics (e.g. you encounter more men on a daily basis or you very rarely go to forests with bears). Or, in other words, just saying "men kill more people than bears" doesn't acknowledge the difference in population sizes between men and bears, and doesn't acknowledge the encounter rate between men and bears. (It may still be that men, per capita, are still more hostile) Of course, this still wasn't an excuse to leave out pro-bear examples and I may consider deleting/changing this video. I did mention that this video wasn't meant to reflect real world numbers but that point can easily be lost while watching the video.
Honestly, it's a pretty genius thought experiment. It reminds me of the "Who's Afraid of Red, Yellow and Blue" paintings by Barnett_Newman. At first, there doesn't seem to be anything special about them (if you're not someone who has an eye for technical brushstrokes), and they just seem like the modern art stereotype which the cynic mocks: "well, even I could have done THAT!" In the end, the abstract painting was so hated that it provoked a knife attack, ripping directly into the artwork. Which rules. Somehow, it makes the title feel ironically prophetic: in its destruction, it fulfilled its artistic purpose, showing exactly who was "afraid" of red, yellow and blue, in the properly "phobic" sense of the word. It was right in the name, and they still fell for it! This thought experiment is much the same. The Man vs. Bear as a hypothetical is a fine enough way to show just how much the stigma against men has grown due to everyday violence, but it truly completes a poetic loop with the reaction of the manosphere towards women picking the bear: from the general sentiment of "but what about how you being afraid of me makes ME feel?!" or "women who picked the bear DESERVE to encounter the bear IRL and get [ ]", these men are doing nothing but proving WHY someone would rather not pick them. You just showcased why, man! You fell for the hypothetical, bro! It's not about all men, even though some people seem to think it IS an attack on the entire male gender. Hell, it's not even about the choice itself. If you would pick the man (or the bear), that's fine. But if you can't UNDERSTAND why someone else would pick the other option, why someone would choose differently, then you may lack empathy in some capacity. And a man of that kind is one of the scariest things to find when you're alone in the woods.
Yeah, and to be completely unbiased, this video also kind of falls for the hypothetical too since the examples I used are "anti-bear". At the time, I thought they made the most sense to use since they appeared to be the easiest to understand/agree on (pro-bear/anti-man cases carry more nuance imo, due to social factors being harder to agree on) but this also likely stems from my bias as pointed out in the last minute of the video. While I do think it's fair to point out that bears aren't as dangerous as some people may believe, I also think it's unfair to say "you are times more likely to be attacked by men" without addressing other relevant statistical data (e.g. you encounter men more frequently). It may be that men, per capita, are still more violent than bears but saying "you are times more likely to be attacked by men" makes men seem worse than they actually are.
Hmmm, undertone feels a bit off. The idea behind the question isn’t to derive some empirical truth: it is to reflect certain sociological sentiments. Also, I think you should be apprehensive in regard to Simpson’s, as you are not generalizing to all men and all bears really: parameters will be distinct. Men in seclusion and men in woods vs. men in society is distinct. And interactions in such distinct scenarios will further be distinct. In fact, I’ll make a bolder claim. I believe you to be misrepresenting some statistical principles. Most importantly, you strawman the argument: bears are not being compared to men. The population of secluded men on trails vs secluded bears on trails are. Plenty of anecdotes exist of both, especially of the latter: bears oftentimes don’t attack without provocation or don’t even realize you’re there (especially black bears: even grizzlies). Then, when it comes to this sampling bias you talk about (more negative experiences with men than bears), I don’t think that’s a real argument against the conclusion. You can say a low trial number (ergo, only interacting with one bear) leads to a rash conclusion, but I think you could group women together to get a satisfactory number, or you could colloquially-void of hard probability-try to use either low trial number or appeal to those who have much greater numbers, and who interact with wild bears for a career or something. Then, to the point of Simpson’s, I think your explanation is a little muddy. What would the covariate be that you’d control for in that example? I think you misrepresented it, and I could be incorrect, but I’d request you walk me through. In this case, the categorical variable appears to be violence across four populations. Where does the association paradox come in? If we analyze men as a group alone, you say that the slope is more extreme than bears when considering only one of the populations, but if you consider both (all men), then it is flatter/similar to bears. If that is your argument, it is a strawman, as stated. The population we worry about is just men on trails. Hopefully that’s coherent enough. Ask for any emphasis.
Ok so I'm gonna use the UC Berkeley Admissions example on the Wikipedia page for Simpson's Paradox (ik, not a great source but gets the job done and Wikipedia maintains a decent level of rigor from peer correctors) Species (Bear/Man) corresponds to Gender (Male/Female) Assault Rate corresponds to Acceptance Rate Location Corresponds to Academic Department I know it's not 1-for-1 but I think it still portrays the idea behind Simpson's Paradox. You say I should be apprehensive with Simpson's Paradox here since I am not noting/making the distinction between groups (poorly summarized but you get the gist) but I believe the entire point behind Simpson's Paradox is that it arises in cases where these factors are not appropriately accounted for (i.e. in situations where statistics aren't properly represented). The first conclusion I reached was that "men are more dangerous than bears" which ignores (as you've stated) the difference between the populations. But if we note the difference between those populations, we can't make that conclusion anymore. So, as I said in the video, the only conclusion we could draw from that particular data set is that walls make bears attack less (again, poorly summarized). Similarly, in the Berekely Gender Gap case, populations of people applying to competitive departments were erroneously combined with people applying to non-competitive departments, resulting in the false claim that admissions were biased in favor of men. In both examples, the differences between populations was ignored and in both examples, a questionable conclusion was drawn. As for the sampling bias point, there seems to be a slight misunderstanding but I do agree that I could've fleshed that point out in more detail. I didn't say the sampling bias existed because women had more negative experiences with men. I meant that people remember their negative experiences more often, meaning that when asked to remember experiences (e.g. male assaultants) assaults will be sampled more frequently than some random people you walked past for half a second. (Here's where I could've gone more in detail) For the average person not trained to deal with a bear, encountering a bear in the wild is usually always negative, regardless of outcome, because most people perceive bears as scary tanks of nature. So experiences with bears will almost always be completely sampled in a survey like this. That's why I only mentioned this sampling bias in experiences with males, because most encounters with males aren't negative, and are likely forgotten. Although, there are other factors at play like rarity/frequency of events (rare events are usually remembered more easily than common occurences) that also influence this sampling. I didn't mean to say that bears are more dangerous than men or that men or more dangerous than bears. Part of the reason why I didn't use real numbers was because that conclusion is irrelevant for the video discussion. It's meant to highlight certain problems that can arise if you're not careful with interpreting statistical data. Like how it's technically correct to say that "you're times more likely to be assaulted by men" even though it ignores the fact that you encounter more men per unit time than bears. Or how statistically speaking, given the temperature of the sun, it's very unlikely for the fusion of hydrogen atoms to occur. But because there are so many hydrogen atoms in the sun, fusion still happens at a large scale. Large in terms of # of interactions and not % of interactions. To address some more specific points: You said I was strawmanning. But as I've explained, that was to emphasize the Simpson's Paradox point, in which, I had to misuse statistics; otherwise, there would be no paradox. You said I was misrepresenting some statistics. I agree, some latter parts of the video were misleading for the sake of brevity. They should have been more fleshed out to avoid confusion. You said to use women involved with bears as a career. But handling bears isn't something the average person is involved in. Since the "debate" originated on social media and hypothesized a scenario in which the average woman had to make a choice, it makes sense to use the experiences of the average woman. I do have to note that this is an incredibly simplified interpretation of an arguably fairly complex question. For instance, there is no real way to classify an encounter. For example, bears can smell really far. Does smelling people (potential prey) count as an encounter (I think we both agree this question is more nuanced but you get the gist)? And some people actively seek out bears for scienc-y stuff. Does a trained person actively looking for bears while staying "unnoticed" (bears can probably still smell you) count as an encounter? Stuff like that.
@@jasonmayo you are again conflating the concept of populations. You are referring to two distinct populations when you say all men and men in the woods; same with all bears and bears in the woods. Simpson’s deals with association issues amongst certain categorical variables. In the case of UCB, the population remains same: comparing women who applied vs men who applied. Then you look at a second categorical variable, the department they applied to, and the association issue becomes apparent: they were applying to departments with lower rates of acceptance. This is not true in the example you’ve brought up, in my eyes. The only categorical variable you are looking at is violence, and you are comparing groups. There isn’t an association error to arise there. We can make faulty empirical claims (such as all men in the wilderness are violent), but that’s an issue with the actual population statistics, and not an issue with association (ergo Simpson’s). There is literally zero reason to mention all men: when you say that, the conceptualization is all men in all contexts. The question is very specific. If you were alone in the woods, would you want to see a bear (which probably won’t bother you), or a male who is alone (whom might commit acts of violence, especially sexual violence).
@@jasonmayo further, I’m blissfully aware of an incorrect framework from flawed statements. But that isn’t the question. It is would you rather interact with this group or the other when alone in the woods. The assumption of statistical evidence is that, per capita, bears bother humans less than humans. Sociological factors contribute to violence against women. That’s the question. It’s, again, more sociological than statistical, but you have self-admittedly no reason to dismiss it-and therefore no real reason to create a video about this topic, especially with sensitivity in mind-if you don’t have any empirical data on it.
(copy pasted your comment with slight tweaks) Simpson’s deals with association issues amongst certain categorical variables. In the case of the video, the population remains same: comparing the women who reported being assaulted by men vs the women who reported being assaulted by bears. Then you look at a second categorical variable, the location they assaulted in, and the association issue becomes apparent: they were assaulted in locations with differing assault rates. Or, if we follow your logic for my example, the only categorical variable in the UCB example is admissions and they are comparing groups that applied to Department A, Department B, etc. There isn't an association error to arise here because these groups are clearly distinct. We can make faulty claims (UCB Department A is biased against women) but that's an issue with the actual population statistics, and not an issue with association. By grouping women assaulted by zoo bears with women assaulted by woods bears, the apparent assault rate drops. By grouping women accepted by competitive departments with women in non-competitive departments, the acceptance rates drop. I do admit I made the extrapolation that said "So you might conclude that men are more deadly than bears" and the reverse of that statement later on. But I did also mention that the only conclusion you could draw was that the locations are different (i.e. zoos have walls, woods don't). Or, in other words, dismissing previous conclusions as false. I know the question is very specific. But the video builds up to that. The initial assumption is as follows: "So let’s assume you ask a bunch of women if they’ve seen and been assaulted by men on hiking trails or at the zoo, and you ask if they’ve seen and been assaulted by bears on hiking trails or at the zoo." Or in other words, not the question. With the final assumption being: "asking women how many men they have encountered outside of areas with a lot of people and how many of those encounters ended up in assault. And you’d do the same with bears." It starts with a misinterpretation of the question because that's how a person reading the question on social media might think about the scenario. A lot of people haven't encountered men while alone in the woods (while a lot of people go camping, many usually aren't going alone), so they'd initially draw from the experience of men as a whole. But drawing from men as a whole isn't quite right, so they'd refine their search and think of men they've encountered in places with less people (presumably places where violent men are more likely to act). And then, they'd be more likely to remember negative individuals in scenarios like that than neutral scenarios, which gives us the sampling bias, since their memories haven't been sampled uniformly. And I never said it was woods men vs all men at the end of the video; it was violent men vs woods men (technically quiet/desolate places men). Because not all woods men are violent criminals (that's not to say violent criminal don't exist either). The only time I've assumed "all men" was at the beginning of the video when I combined the population of zoo men and woods men. Obviously the population of men contains more individuals than people hanging out in woods and zoos but that's fairly pedantic in my eyes. "Note that the numbers presented in this video are inaccurate. They are meant to portray the concepts of statistics and do not accurately reflect real world numbers." The video is meant to portray some of the ways you can misinterpret data in this problem. Therefore, there is some point to making it.
(tl;dr on the bottom) Okay, sorry for the random reply but you can ignore most of my previous comments. I my point of confusion was that you were talking about the question strictly as it was presented, that is, woman encounters man and bear in woods. Then, she considers woods men and woods bears and moves on with her choice. The part where I misread was "Most importantly, you strawman the argument: bears are not being compared to men. The population of secluded men on trails vs secluded bears on trails are." Specifically the part "the argument". Due to the ambiguity of the English language, I didn't realize you were referring to the *original* question and thought you were referring to *my* scenario ("the argument" could refer to the original argument or my argument). So the issue is that my video started in a very tangential manner, or basically, a very "strawman-y" manner (as you've noted). It was never meant to start with the *original* question. I thought that by referencing the original question as a "frame" (i.e. a _rough_ outline) the viewer would assume some of the scenarios listed in the video would only be tangentially related but I guess most people aren't as illiterate as me (I mean that seriously. "Frame" [i.e. a _rigid_ outline] doesn't have the context I thought it had, or at least, it was very poorly emphasized in the video). The initial scenario that I posed was men (as a whole) vs bears (as a whole) which I thought was implied through my lack of distinction in saying "So you might conclude that men are more dangerous than bears" and my act of combining the groups of zoo inhabitants and woods inhabitants. Again, I understand this wasn't the question that was *originally* asked and very much looks like a "strawman" but, in the context of stressing the importance of understanding statistical data, I think the scenario (however unrelated to the original question it might be) does a good job of illustrating how easy it is to make one group (as in all men vs all bears) appear *much* safer than another with some questionable "statistics". I thought by introducing a "basic" example of Simpson's Paradox for the *whole* population of bears and men, people would be able to extrapolate to other, slightly more nuanced, instances of the paradox in the context of the *original* question. Such as: bears (in woods) encountered by regular people vs bears encountered by trained professionals. Or men (not so sketchy) encountered alone near hiking trails vs men, alone, in some random part of the woods (sketchier men). Basically: are we sure that the data set we're thinking about doesn't include uncharacteristically docile/hostile members of the populations we're interested in? I also thought that, as I mentioned in my previous comment, the average person reading this question would probably initially consider _all_ men and _all_ bears they've encountered. Similar to how well-fed crocodiles in a zoo aren't likely to attack their retainers but the average person online would get nervous (watching a video clip) by considering _all_ crocodiles, well-fed or not. So, to me, it made sense to use this tangentially related question as a starting point to build up into the *original* question later in the video (of which, I only address briefly). Because the point was to come up with simple examples for statistical concepts, and simple examples to Simpson's Paradox don't really exist in the *strict* framework of the question. (Although slightly nuanced examples do exist, I thought it would be distracting because there would be like 2-3 things _on top of_ Simpson's Paradox that would need explanation.) So to reiterate: Yes, the context for the Simpson's Paradox was a "strawman" because I had failed to make it clear that I was talking about a slightly different question. And, to be clear, my previous comments were arguing that in context of the "strawman", the scenario I posed is an example of Simpson's Paradox (which, I do realize, is what you claimed from the start). Originally, I thought you used "strawman" in the context of misusing statistics (I thought you said I was misunderstanding Simpson's paradox) and not the entire scenario I presented. I know this this train of logic doesn't make sense; I was barely functioning off of caffeine that day and poorly skimmed through your comment. So I'm sorry for all the confusion this comment chain has created. Also, I do think your use of "strawman" is a bit harsh (like in the tone of your comment), but fair (in the use of the word itself). While it's true that the conclusion drawn from that particular argument was anti-bear, I did, 30 seconds prior to the "strawman", state that the numbers were inaccurate and were only there to portray the concept of statistics (i.e. portray how Simpson's Paradox could manipulate a set of data). But I do understand that the sentiment of the video is glaringly anti-bear which does turn it into a "strawman" in the context of the *original* question. And sorry for putting "strawman" in quotes all the time. I realize that it can seem like I'm saying that my scenario wasn't a "strawman" but every time I type "strawman" it gets the red squiggly underneath, even though it's a word people use, and it really bothers me to not have the quotes there. tldr; We were basically on the same page but due to the lack of my reading comprehension, I thought we were on a different page. Plus, random justification for my video structure that you likely don't care about (based on the previous tone of your comments), but to each their own. One thing isn't going to please everyone. I know that this might not address the concerns of your original comment but I thought it would at least nice to clear up the misunderstanding in the resulting comment chain. Because I sure as hell didn't know what was going on 3 days ago (sleep deprived caffeine brain ftw).
Good stuff!
Glad you liked it! Although, since another commentor pointed this out, do be careful in taking the end result too literally. I avoided using numbers because I want the viewer to draw their own conclusion of the debate while keeping these things in mind. Though I do have to admit, my video is obviously biased against bears.
When I heard about the man vs bear question, I thought the point was that men are *perceived* as more dangerous than bears. If men really are less dangerous than bears, what can men do to change the false perception?
I know this Simpson's man he was Northern Irish & he made the backwards Simpsons significance index thing. that my Northern Irish exame board uses because well we can't use the english inverse version.
I cant believe this is how the general public gets exposed to statistics now lol
Actually its not that bad, but i think its funny that some people genuinely started researching and thinking about statistics because of this
Too bad I couldn't do actual statistics with numbers but addressing the issue with data sampling and interpretation is good enough, I guess.
maan the numbers used for the examples are crazy! pretty cool vid tho
Yeah, the numbers are a bit outlandish because I didn't want people to focus on the numbers; I wanted people to focus on the concept. This video is a biased because it uses mostly anti-bear examples, and while I didn't mean for it to turn out that way, in my (likely biased) opinion, I thought these examples were more agreeable and easier to see the statistical error in.
Some other examples could've included:
Bears encountered at the outskirts of forests (i.e. near urban/foresty borders) might (?) be more aggressive due to a number of potential factors (e.g. desperation for food, being pushed out from their territory, actively seeking out humans, etc.) and NOT a good representation of the overall bear population. Since these bears are closer to human civilization, if these bears are indeed more aggressive, then we would have a Simpson's Paradox-like effect when combining this population of bears with other bear encounters.
Bears encountered by people trained to track/deal with bears will result in fewer accidents than people randomly out in the woods, so including these encounters might unfairly nudge the numbers the other way (in favor of bears). Since trained professionals aren't representative of the average human population.
There are a lot of social/biological factors that impact human-human interactions that could go either way (pro/anti male). These factors probably aren't relevant to bears (you're either food/not food) and if they are, we have no way of accurately checking (we don't speak bear).
What constitues an encounter? Bears can smell very far and might determine that you're food that's not worth chasing after. Does walking past a man on a dimly lit street for 1.5 seconds count as a male encounter? What if that man was across the street? In my opinion, there is a huge gray area in terms of effective encounter range which make it hard to pin on precise and agreeable numbers.
Points like these carry a lot more nuance, which I thought might distract from the statistical principles mentioned in the video. However, this led to my mistake of publishing a video that was anti-bear. When my intention was to shine light to the fact that saying a true statement like "you are times more likely to be attacked by a man than a bear" can potentially be misleading if you don't look at all the relevant statistics (e.g. you encounter more men on a daily basis or you very rarely go to forests with bears). Or, in other words, just saying "men kill more people than bears" doesn't acknowledge the difference in population sizes between men and bears, and doesn't acknowledge the encounter rate between men and bears. (It may still be that men, per capita, are still more hostile)
Of course, this still wasn't an excuse to leave out pro-bear examples and I may consider deleting/changing this video. I did mention that this video wasn't meant to reflect real world numbers but that point can easily be lost while watching the video.
Honestly, it's a pretty genius thought experiment. It reminds me of the "Who's Afraid of Red, Yellow and Blue" paintings by Barnett_Newman. At first, there doesn't seem to be anything special about them (if you're not someone who has an eye for technical brushstrokes), and they just seem like the modern art stereotype which the cynic mocks: "well, even I could have done THAT!"
In the end, the abstract painting was so hated that it provoked a knife attack, ripping directly into the artwork. Which rules. Somehow, it makes the title feel ironically prophetic: in its destruction, it fulfilled its artistic purpose, showing exactly who was "afraid" of red, yellow and blue, in the properly "phobic" sense of the word. It was right in the name, and they still fell for it!
This thought experiment is much the same. The Man vs. Bear as a hypothetical is a fine enough way to show just how much the stigma against men has grown due to everyday violence, but it truly completes a poetic loop with the reaction of the manosphere towards women picking the bear: from the general sentiment of "but what about how you being afraid of me makes ME feel?!" or "women who picked the bear DESERVE to encounter the bear IRL and get [ ]", these men are doing nothing but proving WHY someone would rather not pick them. You just showcased why, man! You fell for the hypothetical, bro!
It's not about all men, even though some people seem to think it IS an attack on the entire male gender. Hell, it's not even about the choice itself. If you would pick the man (or the bear), that's fine. But if you can't UNDERSTAND why someone else would pick the other option, why someone would choose differently, then you may lack empathy in some capacity.
And a man of that kind is one of the scariest things to find when you're alone in the woods.
Yeah, and to be completely unbiased, this video also kind of falls for the hypothetical too since the examples I used are "anti-bear". At the time, I thought they made the most sense to use since they appeared to be the easiest to understand/agree on (pro-bear/anti-man cases carry more nuance imo, due to social factors being harder to agree on) but this also likely stems from my bias as pointed out in the last minute of the video.
While I do think it's fair to point out that bears aren't as dangerous as some people may believe, I also think it's unfair to say "you are times more likely to be attacked by men" without addressing other relevant statistical data (e.g. you encounter men more frequently). It may be that men, per capita, are still more violent than bears but saying "you are times more likely to be attacked by men" makes men seem worse than they actually are.
well i'm not quite sure what to say but i'm the first view this video got
Yeah, I'm not sure why I posted this either. The idea came to me while on the bus home.
@@jasonmayo interesting.
Hmmm, undertone feels a bit off. The idea behind the question isn’t to derive some empirical truth: it is to reflect certain sociological sentiments.
Also, I think you should be apprehensive in regard to Simpson’s, as you are not generalizing to all men and all bears really: parameters will be distinct. Men in seclusion and men in woods vs. men in society is distinct. And interactions in such distinct scenarios will further be distinct.
In fact, I’ll make a bolder claim. I believe you to be misrepresenting some statistical principles. Most importantly, you strawman the argument: bears are not being compared to men. The population of secluded men on trails vs secluded bears on trails are. Plenty of anecdotes exist of both, especially of the latter: bears oftentimes don’t attack without provocation or don’t even realize you’re there (especially black bears: even grizzlies). Then, when it comes to this sampling bias you talk about (more negative experiences with men than bears), I don’t think that’s a real argument against the conclusion. You can say a low trial number (ergo, only interacting with one bear) leads to a rash conclusion, but I think you could group women together to get a satisfactory number, or you could colloquially-void of hard probability-try to use either low trial number or appeal to those who have much greater numbers, and who interact with wild bears for a career or something.
Then, to the point of Simpson’s, I think your explanation is a little muddy. What would the covariate be that you’d control for in that example? I think you misrepresented it, and I could be incorrect, but I’d request you walk me through.
In this case, the categorical variable appears to be violence across four populations. Where does the association paradox come in? If we analyze men as a group alone, you say that the slope is more extreme than bears when considering only one of the populations, but if you consider both (all men), then it is flatter/similar to bears. If that is your argument, it is a strawman, as stated. The population we worry about is just men on trails.
Hopefully that’s coherent enough. Ask for any emphasis.
Ok so I'm gonna use the UC Berkeley Admissions example on the Wikipedia page for Simpson's Paradox (ik, not a great source but gets the job done and Wikipedia maintains a decent level of rigor from peer correctors)
Species (Bear/Man) corresponds to Gender (Male/Female)
Assault Rate corresponds to Acceptance Rate
Location Corresponds to Academic Department
I know it's not 1-for-1 but I think it still portrays the idea behind Simpson's Paradox.
You say I should be apprehensive with Simpson's Paradox here since I am not noting/making the distinction between groups (poorly summarized but you get the gist) but I believe the entire point behind Simpson's Paradox is that it arises in cases where these factors are not appropriately accounted for (i.e. in situations where statistics aren't properly represented). The first conclusion I reached was that "men are more dangerous than bears" which ignores (as you've stated) the difference between the populations. But if we note the difference between those populations, we can't make that conclusion anymore. So, as I said in the video, the only conclusion we could draw from that particular data set is that walls make bears attack less (again, poorly summarized).
Similarly, in the Berekely Gender Gap case, populations of people applying to competitive departments were erroneously combined with people applying to non-competitive departments, resulting in the false claim that admissions were biased in favor of men.
In both examples, the differences between populations was ignored and in both examples, a questionable conclusion was drawn.
As for the sampling bias point, there seems to be a slight misunderstanding but I do agree that I could've fleshed that point out in more detail. I didn't say the sampling bias existed because women had more negative experiences with men. I meant that people remember their negative experiences more often, meaning that when asked to remember experiences (e.g. male assaultants) assaults will be sampled more frequently than some random people you walked past for half a second.
(Here's where I could've gone more in detail) For the average person not trained to deal with a bear, encountering a bear in the wild is usually always negative, regardless of outcome, because most people perceive bears as scary tanks of nature. So experiences with bears will almost always be completely sampled in a survey like this. That's why I only mentioned this sampling bias in experiences with males, because most encounters with males aren't negative, and are likely forgotten. Although, there are other factors at play like rarity/frequency of events (rare events are usually remembered more easily than common occurences) that also influence this sampling.
I didn't mean to say that bears are more dangerous than men or that men or more dangerous than bears. Part of the reason why I didn't use real numbers was because that conclusion is irrelevant for the video discussion. It's meant to highlight certain problems that can arise if you're not careful with interpreting statistical data. Like how it's technically correct to say that "you're times more likely to be assaulted by men" even though it ignores the fact that you encounter more men per unit time than bears. Or how statistically speaking, given the temperature of the sun, it's very unlikely for the fusion of hydrogen atoms to occur. But because there are so many hydrogen atoms in the sun, fusion still happens at a large scale. Large in terms of # of interactions and not % of interactions.
To address some more specific points:
You said I was strawmanning. But as I've explained, that was to emphasize the Simpson's Paradox point, in which, I had to misuse statistics; otherwise, there would be no paradox.
You said I was misrepresenting some statistics. I agree, some latter parts of the video were misleading for the sake of brevity. They should have been more fleshed out to avoid confusion.
You said to use women involved with bears as a career. But handling bears isn't something the average person is involved in. Since the "debate" originated on social media and hypothesized a scenario in which the average woman had to make a choice, it makes sense to use the experiences of the average woman.
I do have to note that this is an incredibly simplified interpretation of an arguably fairly complex question. For instance, there is no real way to classify an encounter. For example, bears can smell really far. Does smelling people (potential prey) count as an encounter (I think we both agree this question is more nuanced but you get the gist)? And some people actively seek out bears for scienc-y stuff. Does a trained person actively looking for bears while staying "unnoticed" (bears can probably still smell you) count as an encounter? Stuff like that.
@@jasonmayo you are again conflating the concept of populations.
You are referring to two distinct populations when you say all men and men in the woods; same with all bears and bears in the woods.
Simpson’s deals with association issues amongst certain categorical variables. In the case of UCB, the population remains same: comparing women who applied vs men who applied. Then you look at a second categorical variable, the department they applied to, and the association issue becomes apparent: they were applying to departments with lower rates of acceptance.
This is not true in the example you’ve brought up, in my eyes. The only categorical variable you are looking at is violence, and you are comparing groups. There isn’t an association error to arise there. We can make faulty empirical claims (such as all men in the wilderness are violent), but that’s an issue with the actual population statistics, and not an issue with association (ergo Simpson’s).
There is literally zero reason to mention all men: when you say that, the conceptualization is all men in all contexts. The question is very specific. If you were alone in the woods, would you want to see a bear (which probably won’t bother you), or a male who is alone (whom might commit acts of violence, especially sexual violence).
@@jasonmayo further, I’m blissfully aware of an incorrect framework from flawed statements. But that isn’t the question.
It is would you rather interact with this group or the other when alone in the woods. The assumption of statistical evidence is that, per capita, bears bother humans less than humans. Sociological factors contribute to violence against women. That’s the question. It’s, again, more sociological than statistical, but you have self-admittedly no reason to dismiss it-and therefore no real reason to create a video about this topic, especially with sensitivity in mind-if you don’t have any empirical data on it.
(copy pasted your comment with slight tweaks)
Simpson’s deals with association issues amongst certain categorical variables. In the case of the video, the population remains same: comparing the women who reported being assaulted by men vs the women who reported being assaulted by bears. Then you look at a second categorical variable, the location they assaulted in, and the association issue becomes apparent: they were assaulted in locations with differing assault rates.
Or, if we follow your logic for my example, the only categorical variable in the UCB example is admissions and they are comparing groups that applied to Department A, Department B, etc. There isn't an association error to arise here because these groups are clearly distinct. We can make faulty claims (UCB Department A is biased against women) but that's an issue with the actual population statistics, and not an issue with association.
By grouping women assaulted by zoo bears with women assaulted by woods bears, the apparent assault rate drops. By grouping women accepted by competitive departments with women in non-competitive departments, the acceptance rates drop.
I do admit I made the extrapolation that said "So you might conclude that men are more deadly than bears" and the reverse of that statement later on. But I did also mention that the only conclusion you could draw was that the locations are different (i.e. zoos have walls, woods don't). Or, in other words, dismissing previous conclusions as false.
I know the question is very specific. But the video builds up to that. The initial assumption is as follows: "So let’s assume you ask a bunch of women if they’ve seen and been assaulted by men on hiking trails or at the zoo, and you ask if they’ve seen and been assaulted by bears on hiking trails or at the zoo." Or in other words, not the question.
With the final assumption being: "asking women how many men they have encountered outside of areas with a lot of people and how many of those encounters ended up in assault. And you’d do the same with bears."
It starts with a misinterpretation of the question because that's how a person reading the question on social media might think about the scenario. A lot of people haven't encountered men while alone in the woods (while a lot of people go camping, many usually aren't going alone), so they'd initially draw from the experience of men as a whole. But drawing from men as a whole isn't quite right, so they'd refine their search and think of men they've encountered in places with less people (presumably places where violent men are more likely to act). And then, they'd be more likely to remember negative individuals in scenarios like that than neutral scenarios, which gives us the sampling bias, since their memories haven't been sampled uniformly.
And I never said it was woods men vs all men at the end of the video; it was violent men vs woods men (technically quiet/desolate places men). Because not all woods men are violent criminals (that's not to say violent criminal don't exist either).
The only time I've assumed "all men" was at the beginning of the video when I combined the population of zoo men and woods men. Obviously the population of men contains more individuals than people hanging out in woods and zoos but that's fairly pedantic in my eyes.
"Note that the numbers presented in this video are inaccurate. They are meant to portray the concepts of statistics and do not accurately reflect real world numbers." The video is meant to portray some of the ways you can misinterpret data in this problem. Therefore, there is some point to making it.
(tl;dr on the bottom)
Okay, sorry for the random reply but you can ignore most of my previous comments. I my point of confusion was that you were talking about the question strictly as it was presented, that is, woman encounters man and bear in woods. Then, she considers woods men and woods bears and moves on with her choice.
The part where I misread was "Most importantly, you strawman the argument: bears are not being compared to men. The population of secluded men on trails vs secluded bears on trails are." Specifically the part "the argument". Due to the ambiguity of the English language, I didn't realize you were referring to the *original* question and thought you were referring to *my* scenario ("the argument" could refer to the original argument or my argument).
So the issue is that my video started in a very tangential manner, or basically, a very "strawman-y" manner (as you've noted). It was never meant to start with the *original* question. I thought that by referencing the original question as a "frame" (i.e. a _rough_ outline) the viewer would assume some of the scenarios listed in the video would only be tangentially related but I guess most people aren't as illiterate as me (I mean that seriously. "Frame" [i.e. a _rigid_ outline] doesn't have the context I thought it had, or at least, it was very poorly emphasized in the video). The initial scenario that I posed was men (as a whole) vs bears (as a whole) which I thought was implied through my lack of distinction in saying "So you might conclude that men are more dangerous than bears" and my act of combining the groups of zoo inhabitants and woods inhabitants. Again, I understand this wasn't the question that was *originally* asked and very much looks like a "strawman" but, in the context of stressing the importance of understanding statistical data, I think the scenario (however unrelated to the original question it might be) does a good job of illustrating how easy it is to make one group (as in all men vs all bears) appear *much* safer than another with some questionable "statistics".
I thought by introducing a "basic" example of Simpson's Paradox for the *whole* population of bears and men, people would be able to extrapolate to other, slightly more nuanced, instances of the paradox in the context of the *original* question. Such as: bears (in woods) encountered by regular people vs bears encountered by trained professionals. Or men (not so sketchy) encountered alone near hiking trails vs men, alone, in some random part of the woods (sketchier men). Basically: are we sure that the data set we're thinking about doesn't include uncharacteristically docile/hostile members of the populations we're interested in?
I also thought that, as I mentioned in my previous comment, the average person reading this question would probably initially consider _all_ men and _all_ bears they've encountered. Similar to how well-fed crocodiles in a zoo aren't likely to attack their retainers but the average person online would get nervous (watching a video clip) by considering _all_ crocodiles, well-fed or not. So, to me, it made sense to use this tangentially related question as a starting point to build up into the *original* question later in the video (of which, I only address briefly). Because the point was to come up with simple examples for statistical concepts, and simple examples to Simpson's Paradox don't really exist in the *strict* framework of the question. (Although slightly nuanced examples do exist, I thought it would be distracting because there would be like 2-3 things _on top of_ Simpson's Paradox that would need explanation.)
So to reiterate: Yes, the context for the Simpson's Paradox was a "strawman" because I had failed to make it clear that I was talking about a slightly different question. And, to be clear, my previous comments were arguing that in context of the "strawman", the scenario I posed is an example of Simpson's Paradox (which, I do realize, is what you claimed from the start).
Originally, I thought you used "strawman" in the context of misusing statistics (I thought you said I was misunderstanding Simpson's paradox) and not the entire scenario I presented. I know this this train of logic doesn't make sense; I was barely functioning off of caffeine that day and poorly skimmed through your comment. So I'm sorry for all the confusion this comment chain has created.
Also, I do think your use of "strawman" is a bit harsh (like in the tone of your comment), but fair (in the use of the word itself). While it's true that the conclusion drawn from that particular argument was anti-bear, I did, 30 seconds prior to the "strawman", state that the numbers were inaccurate and were only there to portray the concept of statistics (i.e. portray how Simpson's Paradox could manipulate a set of data). But I do understand that the sentiment of the video is glaringly anti-bear which does turn it into a "strawman" in the context of the *original* question.
And sorry for putting "strawman" in quotes all the time. I realize that it can seem like I'm saying that my scenario wasn't a "strawman" but every time I type "strawman" it gets the red squiggly underneath, even though it's a word people use, and it really bothers me to not have the quotes there.
tldr; We were basically on the same page but due to the lack of my reading comprehension, I thought we were on a different page. Plus, random justification for my video structure that you likely don't care about (based on the previous tone of your comments), but to each their own. One thing isn't going to please everyone.
I know that this might not address the concerns of your original comment but I thought it would at least nice to clear up the misunderstanding in the resulting comment chain. Because I sure as hell didn't know what was going on 3 days ago (sleep deprived caffeine brain ftw).