2476: Base Rate

Explain xkcd: It's 'cause you're dumb.
(Redirected from 2476)
Jump to: navigation, search
Base Rate
Sure, you can talk about per-capita adjustment, but if you want to solve the problem, it's obvious that this is the group you need to focus on.
Title text: Sure, you can talk about per-capita adjustment, but if you want to solve the problem, it's obvious that this is the group you need to focus on.

Explanation[edit]

Ambox notice.png This explanation may be incomplete or incorrect: Created by a LEFTY. Seems like the entire first paragraph has very little to do with this comic. And there is not enough explanation of why Cueball's statement is a base rate error. Please mention here why this explanation isn't complete. Do NOT delete this tag too soon.
If you can address this issue, please edit the page! Thanks.

The "base rate" is a type of base probability, which a statistical probability can be based on. The base rate fallacy is a type of error in which people are presented with the rate at which something occurs throughout an entire population along with more specific information about a subset of that population, and tend to ignore the whole-population information in favor of the specific information.

In this case, the joke is that 90% of people are right-handed, so if there is no connection between handedness and making base rate errors, then 90% of these errors would be made by right handers. Thus while Cueball's claim that right-handers commit 90% of base-rate errors is technically true, taking that as reason to believe that "making base-rate errors" is somehow specially associated with right-handed-ness -- as would be implied by an intervention effort specific to right-handed-people -- is itself a base-rate error.

Cueball may be holding the pointer in his right hand, suggesting he might be right-handed (as 90% of stick figures are[citation needed]). Since Cueball has no facial features it is impossible to tell if he faces the audience, or looking at his graph. However, it seems most likely that he is looking at his audience while delivering the take home message and thus points at the graph behind him. Thus he likely belongs to the 90% that makes 90% of the base-rate errors, one of those he is just committing.

In the title text, Cueball dismisses the idea of adjusting his graph to account for the difference in numbers of left-handed versus right-handed members of the population. He suggests focusing efforts on the right-handed majority to resolve that 90% of base rate errors. This is a somewhat common counterargument to statistical arguments of this stripe (often as justification for racial profiling, for example); it fails because if the target group is not in fact somehow special with regard to the issue at hand, there is generally "nothing to fix" and no special approach to discover that cannot be just as easily applied to the population of the whole.

Something similar occurs in 1138: Heatmap, where Cueball makes inferences simply based on a population map of the US, instead of statistical evidence.

Transcript[edit]

[Cueball is standing in front of a screen that shows a bar graph with 2 bars with labels beneath. The right bar is significantly higher than the left. Cueball is holding a pointer which he points at the label of the highest bar, which has been encircled.]
Cueball: Remember, right-handed people commit 90% of all base rate errors.
Label: L R


comment.png add a comment! ⋅ comment.png add a topic (use sparingly)! ⋅ Icons-mini-action refresh blue.gif refresh comments!

Discussion

"Aaaand we're back!"
ProphetZarquon (talk) 17:01, 15 June 2021 (UTC)

What happened? Bischoff (talk) 09:02, 16 June 2021 (UTC)
I missed you! 141.101.98.244 19:18, 15 June 2021 (UTC)
There is this : post, but it does not help much yet... But great to be back live as long as it stays like this --Kynde (talk) 12:08, 16 June 2021 (UTC)
I never realised how much I relied on this site to keep me busy until it went down. Beanie talk 10:29, 17 June 2021 (UTC)
903: Extended Mind --172.69.33.111 05:19, 18 June 2021 (UTC)

I can't tell if cueball is holding the pointer in his left or right hand 141.101.98.122 21:30, 15 June 2021 (UTC)

If he's facing the audience, it's in his right hand. Barmar (talk) 21:32, 15 June 2021 (UTC)
Typically cueball has a slightly noticeable 'chin' that indicates the direction he is looking (ex: #2471, #2468, #2460(cell 2 he looks at Megan and cell 3 looks away from her) ). So in this case I'd say he is looking to the right with his body facing the audience. --TallJason (talk) 15:52, 16 June 2021 (UTC)

I wonder if we can expect a comic soon about fan sites going offline. Barmar (talk) 21:32, 15 June 2021 (UTC)

I really don't think Randall keeps an eye on this page... --Kynde (talk) 11:36, 16 June 2021 (UTC)

I removed the claim that Cueball was left-handed; I don't think we can tell whether he is or isn't. --172.70.130.82 22:36, 15 June 2021 (UTC)

There's now a claim that he's right-handed, and I don't think that's reliable either. Yes, the "proper" way to do a presentation is to be facing away from the screen, but I've seen a lot of not-very-good presenters. BunsenH (talk) 04:30, 16 June 2021 (UTC)
I’ll bet that about 90% of the not-very-good presenters you’ve seen were right-handed, therefore if Cueball is not very good, he’s probably right-handed.172.69.34.190 04:53, 16 June 2021 (UTC)

Wouldn't cueball be technically correct, despite his logical error? Given a set of people who make base-rate errors, with no other qualifications, and given that 90 percent are right handed, wouldn't that make 90 percent of the people who make base rate errors right handed? 172.69.71.180 13:21, 16 June 2021 (UTC)

That's the thing about base rate errors; they're not technically wrong, but they are useless from a practical standpoint in doing anything to help the problem, and are misleading about the nature of the issue leading people to make false conclusions, even if there is no detail that is not technically correct on its own. Though 90% of the problem is with right handed people, that is only because 90% of all people are right handed, and trying to incorporate a focus on right handed people in any potential solution is no help at all.--172.70.126.2 23:12, 29 June 2021 (UTC)
The triangle formed between his legs and the floor is isosceles. This would require either that he his stance is perpendicular to the viewer, or else he has non-standard hip geometry. With that premise, if we look at the lines of his arm on the 'right' (from the reader's perspective) the line between his arm and leg are parallel. In order for him to be facing the audience, but yet pointing backwards with his left hand, he would have to either rotate his torso, which would cause the arm/leg lines to be not parallel, or have a missing scapula in his right shoulder (normal human arms can't rotate backwards that far. Alternately, due to the visual placement of the whiteboard, cueball would have to be a very short individual for the base line to show between his biceps, or the board would have to be hung unusually high on the wall. None of that is conclusive, but the most reasonable explanation is that Cueball is left-handed, even if that's bad presentation form.172.69.68.79 15:21, 17 June 2021 (UTC)
Sure, but i think his non standard hip geometry is pretty evident given that his femurs attach directly to his spine. I think I've seen cueball shrug, from which i might infer that he had scapula, but from looking at him, I think it's easily likely that he has a tendon connecting his humeri to his skull. But i do think he's rotating his trunk anyway, because of his chin position. I find irony in the fact that a right - handed person would be making this erroneous analysis. Btw I appreciated this postural analysis so much that i made an account and am posting my ever commentMx. Creant (talk) 05:15, 19 June 2021 (UTC)
Welcome to the community! Please be mindful of the fact that written communication lacks so much context. Your post could be reasonably read to be pretty mean and dismissive in tone. That's how I read it at first. I'm making the conscious choice to re-read it in the best possible light and say thanks for the kind comment re: my post inspiring you to join the community. :) DevAudio (talk) 14:32, 21 June 2021 (UTC)

Can someone explain here in the comments how, in the explanation, we go from that example of 1% / 5% false-positive rate to a 17% / 83%? 172.69.34.190 17:14, 16 June 2021 (UTC)

Take a population of 10,000 tests. From the premise: 1% (100) are true positives and 99% (9,900) are true negatives, regardless of testing. 5% of those TNs (9900*5% = 495) register positive, falsely. We aren't given a false-negative rate, so assuming all 100 TPs register as positive, correctly. 495+100=595 people showing as positive, but only the 100 TPs actually truly were, which is slightly less than 17% (100/595 = 16.8ish%) who have an accurate positive test, leaving a whopping value of slightly more than 83% of tested-positive individuals who were wrongly identified as positive. 162.158.158.105 18:11, 16 June 2021 (UTC)
Another explanation: the 5% is the answer to "What's the change of having a positive test knowning I'm healthy?" the 83% is the answer to "What's the change of being infected knowing I've got a positive test ?".
The latter answers "What's the chance of not being infected, knowing I have a positive test?", as I'm sure you really meant, but that's otherwise an excellent form of summary. 141.101.99.177 13:33, 17 June 2021 (UTC)

Removed the example from the first paragraph. Preserving it here for future reference/improvement. The example is more complex and harder to understand than the example in the comic.

"For instance, imagine a disease that is present in 1% of the population, for which there is a test with a 5% false-positive rate. This test might be presented as "95% accurate", and so people who receive a positive result from such a test are likely to think they have the disease. However, someone who receives a positive test result has only a 17% chance of actually having the disease; a much more likely reason for the positive result, occurring in 83% of all positive test results, is a false (wrong) positive." 172.70.35.69 00:03, 18 June 2021 (UTC)