3104: Tukey

Tukey

|< < Prev #3104 (June 18, 2025) Next > >|

Title text: Numbers can be tricky. On the day of my 110th birthday, I'll be one day younger than John Tukey was on his.

|< < Prev Random Explanation Next > >|

Explanation

The comic gives a quote by John W. Tukey, an American mathematician and statistician, from his paper called The Future of Data Analysis, where he discusses the importance of facing uncertainty. A stripped-down version of his quote could be formulated as:

It is far better to have an approximate answer to the right question than an exact answer to the wrong question.

Randall then gives Tukey's age as 110.000 years (that is, 110 years to three decimal places) — a very precise answer to the question of "how old is Tukey?" However, when it comes to his birthday, arguably the right question in this context, he only gives the approximate answer of "sometime this week". Tukey was born on Wednesday, June 16, 1915, 110 years and 2 days before the release of this comic on Wednesday, June 18, 2025. Since he died in 2000, it can be debated whether he has a birthday this week!

Since the comic was two days (or 0.54% of a year) late compared to Tukey's actual birthday, three decimal places of zero is actually inaccurate on the day of the comic's release, when he was 110.005 years old. However, sometime during that week he was exactly 110.000 years old — the joke being that he gives a very precise age, but not the exact time when this answer was accurate.

The title text states that Randall would be one day younger than Tukey would be on his 110th birthday. Tukey's 110th birthday (on Monday) marked 40,178 days since his birth. Randall's 110th birthday (2094-10-17) will occur 40,177 days after his birth, due to having only passed through 27 leap-days (the first in 1988, the latest in 2092) instead of Tukey's 28 instances (from 1916 to 2024, inclusive).

There are multiple different definitions of the year, however. The one most applied to birthdays is "a whole year has passed when it's the same calendar day", but the usual scientific/mathematical value is the tropical year of 365.2422 days (the time between the same two equinoxes) which smears the leap day across four years (which might impact the last digit of the calculation), and also the sidereal year of 365.2564 days (the time it takes the Sun to return to the same place in the sky), which differs due to precession of the equinoxes. The difference between the tropical and sidereal calculations over 110 years is about 1.5 days, so for this alternative definition of the year, we get a different answer for which day is 110, but it still remains "this week". There's also the average length of a year in the Gregorian calendar (the calendar in use now), which averages 365.2425 days over the 400-year cycle of the calendar. Tukey suggests we shouldn't get hung up on this, but those who happen to be born on 29th February may disagree.

Transcript

This transcript is incomplete: Transcribed by an idiot. Don't remove this notice too soon. If you can fix this issue, edit the page!

[Text formatted as a block quote]

"Far better an approximate answer to the right question, which is often vague, than an exact answer to the wrong question, which can always be made precise."

John W. Tukey

The Future of Data Analysis (1962)

Caption below the comic:

Happy approximate birthday to John Tukey, author of my favorite statistics quote, who was born 110.000 years ago sometime this week.

  Add comment        Create topic (use sparingly)       Refresh 

Discussion

i dont get this comic :( Broseph (talk) 20:42, 18 June 2025 (UTC)\

The main panel makes a joke that the figure of 110,000 years is precise but wildly wrong while that Tukey's birthday is "sometime this week" is vague but basically correct. The alt-text is most likely true (I haven't checked) because of leap years.

I think you'll find it is 110.000, not 110,000 1.146.44.41 23:54, 18 June 2025 (UTC)

Maybe they live in that country mentioned in the alt text of two comics ago. 2601:647:8500:1E09:D4EE:315E:E684:A802 02:13, 19 June 2025 (UTC)

yes {facepalm}. I think I'll find I need to increase font size everywhere yet again so that "." Looks different from "," because I read 110 thousand and not the correct of 110 point 000. Not sure if the joke is different and is funny either way. It's ibuprofen* getting old but Saul Goodman because of the alternative. 2607:FB91:164E:4264:ACA5:3DBE:8045:7E6B 05:55, 19 June 2025 (UTC)

* Spellcheck suggested "ibuprofen." I don't know why or what I typed to. Love you 2607:FB91:164E:4264:ACA5:3DBE:8045:7E6B 06:00, 19 June 2025 (UTC)

yeah I think the joke is that rather than providing an approximate answer to a vague question or am exact answer to a precise question, there's an implied precise question of "exactly how old in years is Tukey", which he answers approximately by adding "sometime this week" to it. In theory he could add as many significant digits passed the decimal as he wants to 110 since it will fall within that one week range no matter what. The actual instant that he turns 110 is of little consequence 2600:4041:2e5:b900:4331:7a73:2a81:46ae (talk) 19:33, 19 June 2025 (UTC) (please sign your comments with ~~~~)

The one day difference is probably because of the rules of leap years. Most century years (like 1800, 1900, 2100) do not have a leap year, but 2000 did have a leap year. Leap year placement is done to approximate Earth's ratio of 365.2422 days per year. Oh, wait. Tukey (1915-2000) and Randall (1984-2094) both lived through the 2000 leap year. So it must just be because Randall was born shortly after Feb 29 of 1984, whereas Tukey was born shortly before Feb 29 of 1916. So Tukey would have had 28 leap days vs. Randall's 27 leap days on their 110 year birthdays. 134.134.139.69 21:09, 18 June 2025 (UTC)

I hadn't read this, before I edited in my version of the explanation (and a few more things surrounding it). Yes, it's basically where the "spans of four years" lie within the whole 110 width. Tukey had one soon (within a year) of his 0th birthday and another just in time (just more than a year) before the 110th birthday. It'd work the same for any year-span that started on the same day on any similar Y mod 4 type of year (1915, also 1911 or 1924), so long as you didn't let the range start before 1900 or finish later than 2100. It gives the same result for 1918+-4n, too, for the same 16th June date in other respects. But shift to the same date in 1916(+-4n) or 1917(+-4n), and it traverses one less leap-day. You can move the date around, of course. If you keep it the right side of the the last/Feb->1st/Mar boundary, as you do for Randall's DOB, then it's still faithfull (1984=1916+4n, where n=17). If you jump back into January or February, it'll become an honourary member of the prior year's thing, but not applicable.

Anyway, Randall's Leap-pattern is two years adrift from Tukeys, which guarantees that his leap-day-count is one different. One way or another. (A one-year mod-difference would half the time be "in the same pairing" and the other half be "in the other pairing".)

Though that only applies for relative ranges that are both entirely within-and-inclusive of 1/Mar/1900 to 28/Feb/2100. You'd have to add another Zeller-like term to the [Y mod 4] thing to 'adjust' if you went further out, and may be able to find two year-ranges that had daycounts two different from each other, I guess, as well as ones that might be the same even though being on mod4+2... But I leave that as an excercise to data-divers wanting to go beyond merely the two indidividuals that the comic specifies. (And don't forget the Julian-to-Gregorian conversion scheme/timing, if you start to encroach upon dates that (for a given locale) are further complicated by a 10-13 day (2, 3, 0 or 1) mod-shuffle. Not including those (e.g. Lithuania, etc) who jumped back out to Julian 'temporarily' again, just to further complicate matters. Ignoring any possibility of the non-400-year manifestation of the 100-year glitch that would warrant a minor additional detail) 92.23.2.228 23:07, 18 June 2025 (UTC)

Note that this is not saying 110,000 (1.1e5) years, but 110.000 (1.1e2), which is in fact the correct number of years. The value has three digits after the decimal point to imply sub-year precision, which is seldom meaningful with birthdays. 2403:5803:BF48:0:0:0:0:1 21:19, 18 June 2025 (UTC)

0.001 years is about 8 hours, so you do need that many digits to be precise to the day. But then he approximates with "sometime this week" -- a week is about 0.02 years. Barmar (talk) 21:33, 18 June 2025 (UTC)

Wow 🤯 no name 02:00, 19 June 2025 (CEST)

One day younger, but exactly the same age in sidereal years (or epochal seconds). Nitpicking (talk) 01:29, 19 June 2025 (UTC)

For a moment there I thought that 110.000 was a binary number, chunked every three digits to make it easier to convert to octal. But the number comes to 48 decimal (60 octal), which is clearly not enough for Tukey's age! --Itub (talk) 11:12, 19 June 2025 (UTC)

Tukey gave us a bit of weird nomenclature for mathematical constructs, such as a cepstrum of quefencies. He caused consternation for editors and proofreaders. (My spell-check thingee wants cepstrum to be strumpet, but I shan't change it.) 173.188.194.66 13:27, 19 June 2025 (UTC)

Two persons born the same day can have an age difference of almost 50 hours, if person A was born at 0:00 on Line Islands and person B at 23:59:59 on Howland Island or Baker Island (the latter two are currently uninhabited, however). Yes, measuring time of birth with second precision is debatable ... --134.102.219.31 14:52, 19 June 2025 (UTC)

Or if one of them was accelerated to close to the speed of light for a sufficiently long journey. 82.13.184.33 15:57, 19 June 2025 (UTC)

One of my favourite XKCD comics yet no name 19:35, 19 June 2025 (EST)

I'm having trouble parsing that Tukey quote. What can always be made more precise? 155.190.35.5 14:58, 23 June 2025 (UTC)

At it's core "It is far better to have 'an approximate answer to the right question' than 'an exact answer to the wrong question'."

The former is being said to be often vague (given that we already know "the answer" is approximate, I'm reading that as "the question"). It is suggested that the latter can always be made made more precise (similarly, by my reading it's not the answer, which is already exact, but the question, which was wrong).

But noting that "precise" and "exact" don't necessarily mean the same thing. Compare precision and accuracy (or possibly vice-versa). It'd be nice to know the surrounding context, as per the speaker's field and conventional use of language, to narrow down what is supposed to mean what... assuming I'm even asking the right question in the first place! 92.23.2.228 17:01, 23 June 2025 (UTC)
        Add comment