Your pet peeve, technical or otherwise.

[DrG]

I have long held a pet peeve against the misuse of the word random, especially in professional contexts, but also everywhere else.

If a number between 1 and 100 is chosen at random, it means to me that every number from 1 to 100 had an equal probability of being chosen.

If a shirt to wear is chosen at random, it means that every shirt had an equal probability of being chosen. If I chose a shirt to wear for some unspecified reason, some unknown or unclear reason or it struck my fancy at the moment, it does not mean that it was chosen at random.

If some woman came up to me and starting talking politics at the mall, it does not mean that it was some random woman.

Making a choice without some specified method or conscious decision does not make it a random choice.

An understanding of what random does and does not mean is an integral part of all inferential statistics that I use and deserves accurate use in many professional activities and in common and casual conversation. It is a very important concept and, as such, it should be treated as a reserved word. Instead, you can read dictionary definitions for the word that are explicitly wrong.

There is no reason to part from the accurate use of the word because it is somehow convenient or one does not know any better and, further, the rationalization of “you know what I mean” is particularly annoying because you don’t know what you mean.

So you are saying that the choices from your selection process do not have a uniform probability distribution they are not random?  Certainly limits whole books full of statistical tools.  And means that many natural processes are not random.  Things like Johnson noise.

No, I am not saying that at all. That is why you could not point to where I said such in my post. It's just that simple and you telling me what I said does not mean that I said it, no matter what you want to believe. Nor does you deciding to interpret what I said in some manner that you can argue about mean that I said it - that is, I don't think what I wrote needs your your interpretation. I stand by what I said, I am not backing down, changing it or otherwise suggesting that it needed to be written more clearly.

What it does mean, however, is that I am not going to take your bait. Nice try and all that but you see it is MY pet peeve and I don't need for you to agree with it, and I am not dismayed that you (or really anyone) disagrees with it, does not understand it, or even verbally engages in precisely what I am peeving about -or not.

If I spent my next 1000 posts trying to convince you otherwise, I feel certain that I would fail because I don't believe that you have tried to understand what I wrote, you just didn't like how it sounded so you decided that I must have said something else and something that you could argue about, hence the distinctly underwhelming "so you are saying" ploy.

If you (or anyone) truly wants to pursue this, then take my post, one sentence at a time, and clearly state why you thing it is wrong. Don't give me this..."so you are saying" and than insert some ridiculous interpretation - you are better than that.

Use what I actually wrote, so:

If a number between 1 and 100 is chosen at random, it means to me that every number from 1 to 100 had an equal probability of being chosen. Do you disagree with that? - if no, move on to the second sentence. If yes, explain to me why what I said in that sentence is wrong or why you don't agree with it even if you don't think it is necessarily wrong.

Can you actually do that? If yes, then do it and if it is a sincere attempt, I will respond.

[PlainName]

If a shirt to wear is chosen at random, it means that every shirt had an equal probability of being chosen.

Not all the shirts may be in the set. You might think "I'm not going to wear a red shirt today" then pick a random one from what remains. The shirt is still chosen at random.

[james_s]

In practice I find that while I have probably 20+ shirts, I end up wearing the same 10 or so shirts over and over, not counting the separate selection of stained/worn out ones I put on when I'm working on something dirty or amusing shirts I might wear to extremely informal events. I start working my way through the shirt drawer and then at some point my partner decides to do laundry and brings me a stack of clean shirts that I put back in the drawer. These end up on top so unless I go out of my way to mix up the stack there is a layer on the bottom of the drawer that I almost never wear.

[Circlotron]

In practice I find that while I have probably 20+ shirts, I end up wearing the same 10 or so shirts over and over, not counting the separate selection of stained/worn out ones I put on when I'm working on something dirty or amusing shirts I might wear to extremely informal events. I start working my way through the shirt drawer and then at some point my partner decides to do laundry and brings me a stack of clean shirts that I put back in the drawer. These end up on top so unless I go out of my way to mix up the stack there is a layer on the bottom of the drawer that I almost never wear.

You could always always take your shirts out and leave them lying all over the floor and wait for your partner to pick them up and put them away again, thus randomising the stack order. I hear that works well.

[james_s]

You could always always take your shirts out and leave them lying all over the floor and wait for your partner to pick them up and put them away again, thus randomising the stack order. I hear that works well.

That sounds like a great way to end up single, at which point I guess I could just wear the same shirt every day until it wears out and then replace it.

[DrG]

If a shirt to wear is chosen at random, it means that every shirt had an equal probability of being chosen.

Not all the shirts may be in the set. You might think "I'm not going to wear a red shirt today" then pick a random one from what remains. The shirt is still chosen at random.

I think you and I have a different concepts of what random means. If you agreed with the prior sentences as per my last post, I will elaborate.

If a shirt to wear is chosen at random, it means that every shirt had an equal probability of being chosen. If I chose a shirt to wear for some unspecified reason, some unknown or unclear reason or it struck my fancy at the moment, it does not mean that it was chosen at random.

I am including the second sentence in that section, even though I note that you are stopping at the first sentence. But I trust that you read the next sentence.

All shirts ARE in the universe of shirts that a person has to wear. Just as in the case of "all numbers between 1 and 100". If I remove one shirt or one number, it is an awkward technicality at best but boils down to rephrasing....ALL shirts except for this red one or all numbers between 1 and 100 except for 64.

In the example case, you now say that, after eliminating a red shirt from possible choices, "The shirt is still chosen at random." when you have presented absolutely no evidence at all that it was chosen at random. Why was it a random choice? What does a random choice mean in your example? it seems like the shirt was chosen and you are deciding it is a random choice because you are saying it is a random choice. It is as though randomness has no meaning at all, only outcome and any outcome is considered to be random. That's ok with me if you believe that, but I don't (and yes I am elaborating and interpreting to some extent but it is to make a point).

I very much disagree and stay with what I said and ask that you read it carefully.

If a shirt to wear is chosen at random, it means that every shirt had an equal probability of being chosen.


This remains coherent and accurate as per following from the first sentence. Because that is, in fact, a [generally accepted] representation of what random means.

Looking forward (not that you are saying it explicitly), it is not a meaningful argument against that sentence to say something like the probability distribution conforms to the list of probabilities associated with each of the possibilities of shirts to wear, so, even if you want to impose, after the fact, that what I am saying is wrong because you decide that there are different probabilities associated with each shirt because it is contrived to the point of being tortured.

We know that different outcomes can have different probabilities but the examples I used are clearly not talking about those distributions and you can tell by what I chose - I didn't say the green shirt has this probability and the white one this probability - they all have the same probability - how could anyone miss that?

Had I said there are 62 red shirts and 128 blue shirts and 4 green shirts and you picked 4 green shirts in 4 choices (without replacement) I could argue that it is not a random choice based on an unlikely outcome and the common theme that each shirt has the same probability of being chosen - but in fact, it could be a random choice no matter how extreme the outcome. That is not the point here.

You know what it means to pick a shirt at random and to pick a shirt for some other reason and SAY something like I chose this shirt at random - that is my point....look at the following sentence...

If I chose a shirt to wear for some unspecified reason, some unknown or unclear reason or it struck my fancy at the moment, it does not mean that it was chosen at random.

...and it most definitely does not mean it was chosen at random because I am stating outright that there was a reason, specified or not, and most definitely that is NOT evidence of a random choice.

Edit: to add two examples: 1) I stand in front of my closet and look at all my shirts and think about the colors and matching with my pants and comfort and and trying to get laid and looking cool or looking for something that goes with a cowboy hat or for some reason choose a shirt. and 2) I stand blindfolded in front of the, arm outstretched and waving back and forth over the totality of my shirts until somebody somewhere (who isn't watching any of this) yells "stop". Then I pick the shirt closest to my outstretched hand.

If used 1 and went to work and said "I chose this shirt at random". I would peeve myself off because I would be misusing the word and disrespecting the process which I hold dear to me.

If I used 2 and went to work and said "I chose this shirt at random". I would not peeve myself off because it was at least approaching a random process.

Note that my peeve is not based on randomness vs. causation, it is simply the MISUSE of the word which underscores the common misunderstanding and under-appreciation of the meaning of the word, in my view.

So, no, it does not look like we agree.

[DrG]

You could always always take your shirts out and leave them lying all over the floor and wait for your partner to pick them up and put them away again, thus randomising the stack order. I hear that works well.

That sounds like a great way to end up single, at which point I guess I could just wear the same shirt every day until it wears out and then replace it.

Well, you may want to save it for some heavy painting or something - I have many such shirts (and pants and shoes), all ready to go when I tackle that dirty job.

[james_s]

Well, you may want to save it for some heavy painting or something - I have many such shirts (and pants and shoes), all ready to go when I tackle that dirty job.

So do I, and I make new ones more frequently than I'd like to, every time I forget to change into grubby clothes, or think "I'm just going to do this quick thing, I'll be careful not to get dirty."

[DrG]

Well, you may want to save it for some heavy painting or something - I have many such shirts (and pants and shoes), all ready to go when I tackle that dirty job.

So do I, and I make new ones more frequently than I'd like to, every time I forget to change into grubby clothes, or think "I'm just going to do this quick thing, I'll be careful not to get dirty."

I favor the bleach screw up more than anything....I just need to use a little bleach here and I will be real careful...instant grey shirt with added pink streaks - it truly amazes me.

[PlainName]

I think you and I have a different concepts of what random means.

That seems to be the crux of it, yes.

Which is fine. But it seems to me that your peeve is based on your preferred meaning of the word, which is not the commonly understood meaning. Kind of like me complaining that everyone in the US calls the pavement a 'sidewalk'.

[themadhippy]

calls the pavement a 'sidewalk'.

which actually describes what it is, a walk along side,same with boot and trunk,trunk is were the trunks (luggage) goes,just a pity they keep leaving letters out of words.

[mc172]

calls the pavement a 'sidewalk'.

which actually describes what it is, a walk along side,same with boot and trunk,trunk is were the trunks (luggage) goes,just a pity they keep leaving letters out of words.

"Burglarized". I'll leave it there.

[themadhippy]

"Burglarized". I'll leave it there.

Indeed, we all know the correct spelling is BurglariSed

[IDEngineer]

calls the pavement a 'sidewalk'.

which actually describes what it is, a walk along side,same with boot and trunk,trunk is were the trunks (luggage) goes

Consider: In the USA we drive on the parkway, yet park on the driveway.       

[coppice]

My pet is peeved, because I haven't taken her to the park for a few days. She feels she is leading a dog's life.

[Sal Ammoniac]

Counter-peeve: People who do not understand 600 years of typesetting evolution conclusions.   

One of my personal peeves: using sans-serif fonts for body text.

[TimFox]

Some examples of the statistical properties of random processes.  In each of these cases, the next result/event/choice is independent of its predecessors.  The first one shows a flat distribution, where each outcome has the same probability, but the others do not.  Having equal probabilities for each outcome, in general, is neither necessary nor sufficient to define a random process, but is not excluded.

1.   Honest die.
An honest die (plural dice) is a cube with six faces numbered 1 to 6 (opposite sides add to 7).  The probability of any number from 1 to 6 equals 1/6.  The mean value for the numerical value of the events (throws) is 3.5, which is not itself a possible value.  The events are independent, so that the probability of each of the six results after throwing a “4” is the same as after throwing a “3”.  This is close to randomly choosing shirts that have laundry markers to distinguish them, assuming you have the same set of shirts available for each act of random choice.  If used shirts stay in the laundry hamper for a long time, there will be an obvious change in the probability distribution as the number of available shirts decreases, but the distribution will be “flat” for each act.

2.   Loaded die.
If a malefactor drills a small hole through the single spot on the “1” side, inserts a heavy ball, and re-seals the hole with the ball near the “1” face, the probabilities are no longer equal:  “6” is more probable than “1”, but “2” through “5” have equal probabilities.  The process is still random when throwing many times, but the probability distribution is no longer “flat”.

3.   Astragalos.
The astragaloi (Greek) or tali (Latin) were six-sided ankle bones used as dice in ancient Rome.  They had six sides, but two were curved and did not count, leaving four sides (scored by which side ended down) that were numbered by convention as I, III, IV, and VI.  Since the bones are not cubical, the probability for each of the four sides is different.  The probabilities add up to 1.

4.   Poisson statistics.
This is very important in physics and engineering for the probability of a given number of events occurring in a pre-defined time interval.  The usual example is a radioactive source and a counter that registers the number of detected particles in a fixed time.  See any statistics text or Wikipedia for details.  Specifically, assume that the mean rate (per that time interval) is the parameter m, and let x be the random variable of detected particles in each time interval, which can be any non-negative integer k = 0, 1, 2, 3 …  As is well known in statistics, the standard deviation of the distribution is given by m^1/2.  The probability distribution is far from flat, and shows a pronounced peak.

The probability that x = k for each time interval is given by the formula

Pr(x = k)  =  m^k exp(-m)/k!

For m = 4, as an example, we have these probabilities (calculated in Excel) which go on forever, but die off quickly:

(k = 0)  0.018
(k = 1)  0.073
(k = 2)  0.147
(k = 3)  0.195
(k = 4)  0.195
(k = 5)  0.156
(k = 6)  0.104
(k = 7)  0.060
(k = 8 )  0.030
(k = 9)  0.013
   ….
The probabilities for k = 3 and k = 4 are equal, even though the mean value is 4.

[mansaxel]

For a truly fascinating story of an early attempt at doing random numbers electronically, see

https://datamuseum.dk/w/images/5/56/DASK_daskrnd_en.pdf

which is an off-print of Teleteknik, Engelsk Edition, Vol III, 1959,No 2: A Generator of Random numbers. by Henning Isaksson. Thanks to forum member "bsdphk" who works with older Danish technology history and IIRC put me on the trail of that text. (In true rabbit hole fashion, I then spent 3 days reading about Regnecentralen and other early Danish computer history. You've been warned. Most of it is in Danish, though, so it helps that I'm a passable reader and speaker of that language...)

[james_s]

I favor the bleach screw up more than anything....I just need to use a little bleach here and I will be real careful...instant grey shirt with added pink streaks - it truly amazes me.

Oh yeah, that stupid Chlorox Cleanup stuff, I've ruined more clothes with the spray bottle of that. "Oh I'll just give this spot of mold mildew in the shower a quick squirt", and then the mist blows back at me and makes little pink spots all over whatever I'm wearing. Or I wipe down the counter with it then forget and lean over it to grab something or drag my sleeve across it. I've got shirts, hoodies, pants, even pajamas and boxer shorts that have bleach streaks or mist spots on them. Sure does a good job of keeping the shower clean though.

[james_s]

Which is fine. But it seems to me that your peeve is based on your preferred meaning of the word, which is not the commonly understood meaning. Kind of like me complaining that everyone in the US calls the pavement a 'sidewalk'.

To me anything that is paved is "pavement", that includes both the street and the sidewalk, parking lots, etc. The pavement is the paved surface, a sidewalk is a paved surface for pedestrian traffic adjacent to a road.

[AaronLee]

Counter-peeve: People who do not understand 600 years of typesetting evolution conclusions.   

One of my personal peeves: using sans-serif fonts for body text.

Then it must pain you to read EEVblog Forum, with it's standard sans-serif font.

Actually I worked for many years on a project where fonts were one of the top priorities. Many people think they don't like a sans-serif font for normal text, but in cases where it's hard to read (the size of text is small relative to the reader's distance from the monitor), 99% of the time, in my experience, with countless people in my dataset, they agree that a sans-serif font is more readable. If you're sitting in front of your computer monitor and can adjust the font size yourself, then a font that you find attractive is fine. If your application involves reading from a distance, where it's not practical to increase the font size anymore, choosing a font that is readable becomes the priority over aesthetics.

For someone like myself, getting older and having poorer eyes, I prefer a sans-serif font even for regular stuff on my computer monitor. When stuff appears in the normal default font size, if it's in a sans-serif font, there's less likelihood I'll need to increase the magnification to be able to easily read it.

[Sal Ammoniac]

Counter-peeve: People who do not understand 600 years of typesetting evolution conclusions.   

One of my personal peeves: using sans-serif fonts for body text.

Then it must pain you to read EEVblog Forum, with it's standard sans-serif font.

Not really. I was thinking about printed books. And it's not just aesthetics. I find that my reading speed is higher with serif fonts and I have less eye strain as well. I read a lot--three or four books a week--so others who read less might not find it as much of an issue.

I also prefer traditional fonts, such as Times Roman, over new-age crap that millennials find edgy and modern.

Another peeve: people who don't understand the difference between "its" and "it's". 

[IDEngineer]

Disclaimer: I used to work for a printer company, and design fonts, in a past life.

Long-past studies did prove that serifed fonts were more readable. However, I saw a decline in that in the 80's and 90's. I believe it was due to the relatively low resolutions which were available then (think dot matrix printers, low-res CRT's, etc.). The long-past studies were done with offset printing and other inherently high resolution printing systems so resolution wasn't an issue. But when you try to render serifed fonts in a low resolution environment, all that extra detail goes from "helpful optical hints" to "blurry smudges".

In a low res environment, simplicity helps - and thus during this time of my career we found that sans serifed fonts were actually more readable and preferred by most customers. The old time print shop owners thought we were nuts, but it wasn't an apples to apples comparison. Neither side was "wrong", the underlying technology was just different and the optimal font was different too.

Today, with laser and ink jet printers regularly hitting actual 600-1200 DPI and LCD/OLED displays having inter-pixel spacing below the eye's resolving power at normal viewing distances, sans serifed fonts are no longer a disadvantage. You can use serifed fonts and resolve every extra detail. However, I must say that all these years later I still prefer sans serif fonts... it's just "less optical noise" while conveying the same information to my eyes. Maybe it's all those years of writing font editing tools and creating font character sets. Who knows.

[AaronLee]

Long-past studies did prove that serifed fonts were more readable. However, I saw a decline in that in the 80's and 90's. I believe it was due to the relatively low resolutions which were available then (think dot matrix printers, low-res CRT's, etc.). The long-past studies were done with offset printing and other inherently high resolution printing systems so resolution wasn't an issue. But when you try to render serifed fonts in a low resolution environment, all that extra detail goes from "helpful optical hints" to "blurry smudges".

Yes, and for my personal experience, it was with fonts on computer monitors during the 80's and 90's. Starting with very low resolution of I think 320x192, then to standard VGA, then SVGA. Often the text wasn't even displayed on a computer monitor, but on NTSC/PAL video monitors which were already fuzzy looking without adding even more fuzziness via serifed fonts. The fonts were all custom hand-drawn and the sizes considerably larger than the built-in ROM character set, so adding serifs were possible, but they still looked lousy.

I don't really read printed books these days, except for the odd opening up a reference book or old text book to just read a section of it. If I was doing a lot of reading of printed books, and if the text size was large enough for my eyes, I'd likely prefer serified fonts for that case. Any of the newer stylized fonts being created these days don't interest me in the least.

[CatalinaWOW]

I have long held a pet peeve against the misuse of the word random, especially in professional contexts, but also everywhere else.

If a number between 1 and 100 is chosen at random, it means to me that every number from 1 to 100 had an equal probability of being chosen.

If a shirt to wear is chosen at random, it means that every shirt had an equal probability of being chosen. If I chose a shirt to wear for some unspecified reason, some unknown or unclear reason or it struck my fancy at the moment, it does not mean that it was chosen at random.

If some woman came up to me and starting talking politics at the mall, it does not mean that it was some random woman.

Making a choice without some specified method or conscious decision does not make it a random choice.

An understanding of what random does and does not mean is an integral part of all inferential statistics that I use and deserves accurate use in many professional activities and in common and casual conversation. It is a very important concept and, as such, it should be treated as a reserved word. Instead, you can read dictionary definitions for the word that are explicitly wrong.

There is no reason to part from the accurate use of the word because it is somehow convenient or one does not know any better and, further, the rationalization of “you know what I mean” is particularly annoying because you don’t know what you mean.

So you are saying that the choices from your selection process do not have a uniform probability distribution they are not random?  Certainly limits whole books full of statistical tools.  And means that many natural processes are not random.  Things like Johnson noise.

No, I am not saying that at all. That is why you could not point to where I said such in my post. It's just that simple and you telling me what I said does not mean that I said it, no matter what you want to believe. Nor does you deciding to interpret what I said in some manner that you can argue about mean that I said it - that is, I don't think what I wrote needs your your interpretation. I stand by what I said, I am not backing down, changing it or otherwise suggesting that it needed to be written more clearly.

What it does mean, however, is that I am not going to take your bait. Nice try and all that but you see it is MY pet peeve and I don't need for you to agree with it, and I am not dismayed that you (or really anyone) disagrees with it, does not understand it, or even verbally engages in precisely what I am peeving about -or not.

If I spent my next 1000 posts trying to convince you otherwise, I feel certain that I would fail because I don't believe that you have tried to understand what I wrote, you just didn't like how it sounded so you decided that I must have said something else and something that you could argue about, hence the distinctly underwhelming "so you are saying" ploy.

If you (or anyone) truly wants to pursue this, then take my post, one sentence at a time, and clearly state why you thing it is wrong. Don't give me this..."so you are saying" and than insert some ridiculous interpretation - you are better than that.

Use what I actually wrote, so:

If a number between 1 and 100 is chosen at random, it means to me that every number from 1 to 100 had an equal probability of being chosen. Do you disagree with that? - if no, move on to the second sentence. If yes, explain to me why what I said in that sentence is wrong or why you don't agree with it even if you don't think it is necessarily wrong.

Can you actually do that? If yes, then do it and if it is a sincere attempt, I will respond.

First, I did over interpret your statement. 

Second, you have made it obvious this is a serious peeve for you. 

I am still not in total agreement.  The disagreement lies solely in how language is used in edge cases or grey areas, not in the fundamental math.  Some examples of where this might apply.

1.  In your case of selection of integers between 1 and 100.  People are notoriously poor as random number generators.  If you conduct a test with hundreds of people asked to select a digit randomly in the interval the results clearly don't follow a uniform distribution.  But there would be no conscious intent at selection, just a poor quality generator.  In fact, armed with the results of the experiment you could even modify the results to better approximate a desired distribution (This is just an application of Bayes Theorem).  I find nothing wrong with describing the peoples choice as random.  The problem comes from assuming more information than actually exists in that word.

2.  The second example is actually conceptually similar.  Use of the flawed random number generator provided with Excel (or any number of other languages).  In this case there is usually at least a partial definition, usually limited to saying a uniform distribution over the interval.  The generator passes simple tests of randomness, but fails a full examination.   But the results are close enough to random that for the vast majority of applications there is no significant error introduced by these failures.  Again I don't really have a problem with casually calling this a random number generator.

3.  In the shirt selection example there may indeed be many biases in the selection process ranging from color and style preferences through difficulty of access in the storage system.  The results may vary depending on whether one looks at the shirts and "tries" to pick at "random" or closes ones eyes and fumbles for a shirt.  Both processes have biases, but the conscious intent is still random, and usually nothing critical depends on the quality of the selection.

I guess my position boils down to a knowledge thing.  If you are doing something that requires high precision in the use of the word random you should also have the knowledge and experience to realize the pitfalls in obtaining axiomatically correct results.  There is no more harm in the word being used casually in general use than there is in saying that your mass is 80 kg based on an uncalibrated bathroom scale. 

Now if a guy is going to be executed if after a year his choice of shirts does not pass formal tests for randomness he better do some serious thought about the selection method.  With stakes that high just numbering the shirts and using one of the published tables of random digits might not be good enough.  He would be well advised to spend time understanding the failure criteria, rules for dealing with shirts lost in the laundry and any new shirts purchased and myriad of other details.  And might well protest on the error bounds of a test based on only 365 samples.