Talk:2696: Precision vs Accuracy

Explain xkcd: It's 'cause you're dumb.
Jump to: navigation, search

87.532% of all statistics are just made up. 11:10, 9 November 2022 (UTC)

Why is 'Barack Obama is 6'1"' and 'Barack Obama has 4 legs' medium precision? It seems to give exact value, so high precision. Tkopec (talk) 11:44, 9 November 2022 (UTC)

OK, I get it. 6'1" means something between 6'0.50" and 6'1.49". For height it's OK, but when counting legs, it seems like a stretch. Tkopec (talk) 12:30, 9 November 2022 (UTC)
The four legs are probably considered to be only medium precise, not because of the number but because of the imprecise term "leg". While humans can walk on all four extremities, thereby using them as legs, the upper two are commonly referred to as arms. Bischoff (talk) 14:54, 9 November 2022 (UTC)
(ECed by Bischoff) Plus a person's height (excluding differences to footwear and perhaps hairstyle) varies by an inch or so over the course of a day, as the spine compresses whilst mostly upright (would depend a bit upon your daily activities, but "an inch" or 2-3cm is the typical quoted value, with all the questions about precision as well as accuracy). Within an inch of such a foot-and-inch value is basically between slightly over a percentage point of drift across a continuum of ultimately non-integer values.
The number of legs is generally a whole number (perhaps lower-limb amputees could claim "half a leg", but is that for above the knee or below or... that's beyond my wish to define, I would leave it up to the individual amputee to finesse to their own liking) and assigning decimals, even .000(recurring), would be over-precise. A definite plain figure (however inaccurate) being the happy and acceptable medium between that and the vague imprecision (never mind inaccuracy) of the kind in the cell below. 15:00, 9 November 2022 (UTC)
The medium is because it says most, and not all! --Kynde (talk) 08:08, 10 November 2022 (UTC)
It says "most cats", indeed, but the above was about Obama, singular. Though I think it's covered anyway... 09:44, 10 November 2022 (UTC)
All the statements about 'Barack Obama' ought to be medium precision at best, because there could be more than one Barack Obama, and it doesn't give any further contextualisation to identify, for example 'the Barack Obama who was president of the United States of America'. 09:29, 11 November 2022 (UTC)
Compare with 6'1"1/50 or 4.0000 legs, both of which would imply a higher degree of certainty.-- 08:58, 13 November 2022 (UTC)

Someone should add an explanation of the difference between precision and accuracy. Nutster (talk) 13:13, 9 November 2022 (UTC)

Tried it myself. Maybe made it too compact, but I often go on too long so I tried made it as brief and snappy as I felt I could. Over to other editors to rewrite or replace. 15:00, 9 November 2022 (UTC)
That there is confusion over this was a bit of a surprise to me, about 20 years ago, when I worked (as I did for many years) in the outdoor pursuits trade. GPS units would give a 12-character grid reference (1m2), but couldn't be relied upon to that level. I would tell people they're more precise than they are accurate, until it became apparent that they were waiting for me to complete the joke they thought I'd begun, as I was so clearly contradicting myself, what with the two words meaning identical things.
Having gone on to explain the difference between the words, the neat brevity I'd sought was lost.
Obviously they can be used sort of interchangeably in casual conversation, but I thought the difference was well enough known that, when talking about a navigational instrument, it would be obvious what was meant.
Nope. Yorkshire Pudding (talk) 20:18, 9 November 2022 (UTC)
I deal with OS Grid References a lot, in a similar context, and a number of people who give 10-digits or more (2x5, for 1m res) from devices that typically don't ever settle down to less than 3m, and provably can be tens of metres off if there happens to be a small tree or shrub nearby.
(In fact, the other day I was geohashing myself, and my device was insisting I was in a totally different bit of the open field, 50m or so, no matter how much I sat it down at the provably correct point and wandered away so that even I wasn't obscuring its view of the sky. But it was good enough for me, which was all I do it for, so after giving it 5 minutes I counted it as done.)
And, in yet another activity, the publicised information for an event included a 12ish-DP reference for the starting area (vaguer than that), but just the postcode for the HQ (a very definite building that you could bullseye on a map), in a rural area where it covered half the valley! 22:19, 9 November 2022 (UTC)

How is 17.082 palindromic? Barmar (talk) 14:54, 9 November 2022 (UTC)

My error, I meant an anagram! (Was going for "anagramic", and my brain clearly rebelled.) 15:00, 9 November 2022 (UTC)

High Precision High accuracy, Randall Munroe misses when Obama was president. Low precision Medium-rare accuracy, so do we, Randall, so do we. (talk) (please sign your comments with ~~~~)

It is so annoying that the US uses . and , to mean the opposite of what most European countries (including Denmark where I live). So when I read this it states that Obama was president less than 3 days (70 hours) but it more than 70000 feet tall. :-) Of course I now the difference but I have to think about it more than if everyone used the same standard. Also height should use SI units as everyone should ;-) (weight given in number of cats is the new SI unit as far as I know, but don't use inches and feet ;-D ) --Kynde (talk) 08:17, 10 November 2022 (UTC)

Well, as a UKian, I was happy enough. Tell you what, though, let's develop a new and mutually-acceptable standard notation... ;) 09:44, 10 November 2022 (UTC)
Good idea. Lets meet on 11/12/22 to discuss the details. --Lupo (talk) 13:41, 10 November 2022 (UTC)

I think Randall missed an opportunity to clarify how high precision can make something inaccurate. He could have said that Obama is 6’ 1.02173” tall, which would clearly be very precise, and also clearly inaccurate, simply because of the excessive precision. John (talk) 15:22, 10 November 2022 (UTC)

Saying 6'1.0278 would have been more in theme, there. And it would be not really more inaccurate (might even be closer to the truth...) but would convey a false precision.
Interstingly, when Andrew Waugh measured Mount Everest (before it was so named) he got a diffraction-adjusted figure of 29,000 feet, but decided to announced that it was 29,002 so that it didn't just like a rough figure rounded to the nearest hundred or even thousand feet. This made him the first person to put two feet on the top of Everest!
(...The actual error was not bad, given his measurements had to be made from hundreds of miles away. Current official measurements with on-the-spot modern GPS say 29,031.7 feet (for the snow-peak, which is all that Waugh could mention), after 170ish years of (by some estimates, but contested) about a foot of extra height per decade through the continuing techtonic raising of the Himalaya. And any unknown differences in snow-depth. Certainly it was within tens of feet, i.e. a dozen or so metres. With a bit of an error-bar, but not really that big when you consider it...)
So, arguably, that case was a deliberately false accuracy to help convey the true precision. 16:15, 10 November 2022 (UTC)
I don't get your point? Unless you just made up everything after the decimal point: How would it be less acurate? --Lupo (talk) 09:37, 11 November 2022 (UTC)
The only thing I can imagine is, that these kinds of numbers happen due to conversions. E.g. 6ft1in would be 185.42cm (according to the first calculator I found), but it is unlikely that 6ft1in was as precise as a cm-value with 2 digits after the decimal point would be. And in the other direction 185cm (which would be the usual precision of a height in m or cm - while 186cm could still be correct as it would be 6ft1in in the "usual precision") would calculate as 6ft and 0.83in --Lupo (talk) 10:18, 11 November 2022 (UTC)
If Obama's height is provided with this much precision, you can assume that the numbers are made up. 0.0278 inches are - in real measure units ;-) - 0.07mm. That's the diameter of a strain of hair. Nobody's height gets measured to that kind of precision. Kimmerin (talk) 08:10, 17 November 2022 (UTC)

I'm not sure the current explanation's claim that 'being too precise usually decreases accuracy' is, er, accurate (or perhaps it's just imprecise). It might be reasonable to claim that increasing precision tends to decrease accuracy relative to the level of precision, but not so much in absolute terms, or even necessarily relative to the size of the thing being measured. 09:38, 11 November 2022 (UTC)

I think it's badly phrased. The assumed accuracy can be degraded and disadvantageous.
For example, to use someone's figures from just above, looking for an individual with a height of 185.42cm might seem to rule out the one that you find is 185.57cm tall, though they are indeed the one initially measured/estimated at 6'1" and would definitely be within an inch or so in this latest attempt to match them.
An old phrase that I grew up with is "don't try to be accurate over inaccurate details" (courtesy of a chemistry teacher, where we frequently used mmol-like measurements in analyses like titrations). The number of articles that say "the probe flew past the asteroid at a distance of about 20 miles (32.187 kilometres) ...", where clearly the accuracy is misleading, especially if the conversion ends up being back-converted by someone else with no idea ("...which is 20.0000746 miles"), and may have come from an original figure actually deliberately pegged at 35km (21.748 miles!), within a few metres or less.
Really, you should be taking the level of precision/accuracy inherent in the initial values, preserving the awkward fractions throughout the intermediate steps and converting the inherent ranges by the same process then clearly presenting the final figure to no more exactitude than the initial smudge of "all actual values that would be given by this type of input value", and maybe less. The write-up might be then be realistically "...of around 21¾ miles (35km)", if using a better primary source, or "20 miles (~30km)" in a case of the detail already being likely lost by intermediate chinese-whispers.
But this is what confuses people. And how even those that are not confused can confuse others... 12:16, 11 November 2022 (UTC)
It gets even better when different units also use different 0s. So for a persons height we can assume that as 0ft0in and 0cm is the same, 185cm is one order of magnitude more precise than 6ft1in, as it is 3 significant digits vs 2 at the same height. However a persons body temperature in 38°C with 2 significant digits and 311K with 3 is the same level of precision and only .15°C (Or .15K) apart, while 100°F (37.77...°C) is also very close but a bit more precise. --Lupo (talk) 14:10, 11 November 2022 (UTC)
One of many reasons that Celsius and Fahrenheit are not considered as true units - their connection to kelvins is affine, not linear. 05:49, 13 November 2022 (UTC)
Acknowledging that Celsius degrees equal Kelvin degrees, which remains a useful equivalence, even though degrees Celsius does not equal degrees Kelvin. (Ditto with Fahrenheit and Rankine.)
...and I'm partial to Delisle, anyway. ;) 11:28, 13 November 2022 (UTC)

I was expecting maybe a reference to Schrödinger's President when I first read the comic - but later realized that this could have been misconstrued as a threat. Oops!

As far as I recall, isn't the transcript supposed to avoid tables? I understand blind people with text reading programs use the transcripts to follow this comic, and thus it should avoid visual elements wherever possible? NiceGuy1 (talk) 04:49, 12 November 2022 (UTC)

Generally, yes, though some useful additional description went in before I might have 'flattened' the description again, and there are ther extant table-transcripts
Best practice would be to not rely on screen-readers to say nice informative things about tabulation and instead say it all explicitly (like they can't be relied on parsing MathML stuff), but there's good manual description and bad, too. 13:13, 12 November 2022 (UTC)

In the fewer-legs-than-your-cat category, any interest in adding a link to the "How many legs does a dog have if you call the tail a leg?" riddle often attributed to Lincoln? The best link I found is which makes it clear the riddle was already in circulation by 1825, well before Lincoln's usage. 05:30, 13 November 2022 (UTC)

If you only look at the squares that mention cats, the resulting shape resembles a basic glider from Conway's Game of Life. I don't know how much this has to do with the comic but it feels like the sort of nerd shit that is worth mentioning. 13:03, 12 October 2023 (UTC)

I came to this page to mention the fact only to find that I already did. 13:07, 21 February 2024 (UTC)

most obamas have hundreds of legs in their cat Squishmallow fan (talk) 21:48, 10 February 2024 (UTC)

INSANELY LOW PRECISION/UNFATHOMABLY LOW ACCURACY Psychoticpotato (talk) 21:40, 13 May 2024 (UTC)
cat leg is hundreds of obama Squishmallow fan (talk) 19:03, 3 July 2024 (UTC)