Looking for a lost comment

June 5, 2015 7:11 AM

I give up. I can't find a really awesome comment which I believe was in Ask.me about costuming for television and women's undergarments. The gist of the comment was rich, good girls or women characters with otherwise desirable traits were Spanx'd, Hollywood-taped and underlined to within an inch of their lives while poor, bad, slutty girls or women characters with otherwise undesirable traits were left without decent foundation garments. Periodically I scour my favorites for it and search for it through the google, but I can't pull it up again. Did *you* say it? Do you remember it? Did I imagine it or read it somewhere else? The bra conversation on the front page finally pushed me over the edge to asking the community if someone else can put a finger on it.
posted by crush-onastick to MetaFilter-Related at 7:11 AM (16 comments total)

Is it this?

The other thing, that I think is huge, is something I learned from a friend who owns a costume warehouse for movie/TV productions. He says that when they want to show "class", the wealthy or good-girl characters get proper undergarments so there are no bulges and no jiggling. Sometimes three pairs of Spanx at one time.

posted by specialagentwebb at 7:26 AM on June 5, 2015 [11 favorites]

Yes! Ha, landspeed records yet again. Thanks. Slightly less detail than I remember on the costuming angle, but more on the clothing as signifier generally.

Thanks,
posted by crush-onastick at 7:33 AM on June 5, 2015 [1 favorite]

I would love to see a statistical analysis on how quickly questions are sometimes answered around here, ranked from quickest on down. I swear there are a large number of AskMe questions that are answered in about three minutes or less on some obscure thing no one has ever heard of (except that one person who just happened to be reading), and it would be cool to see how many times that's actually happened. I've had at least one question answered so quickly after weeks of scouring the web that it made my head spin a little bit.
posted by SpacemanStix at 8:49 AM on June 5, 2015 [1 favorite]

Of course, there's no way to automate that kind of analysis, so I'm not sure how realistic it would be. It couldn't be a "best answer" flag, but an actual "thanks, that's it!" response, which would have to be by hand.
posted by SpacemanStix at 8:54 AM on June 5, 2015 [1 favorite]

Yeah, doing it for Metatalk would be tricky; I doubt there's any good process that wouldn't be totally manual, not least because the first step would have to be identifying threads that are "what/where was this..." questions vs. other types of Metatalks.

With stuff on Ask you could at least take a stab at an automated analysis; look at each question with at least one best answer and compare the timestamp of the question to the timestamp of the first best answer. It'd be far from perfect but it'd be something.
posted by cortex (staff) at 9:02 AM on June 5, 2015

I was thinking primarily AskMe, although it sometimes happens here, too. AskMe regularly blows my mind at how effective it can be. It would be cool to have some sort of analysis/infographic of that. I wouldn't be the guy to set up any sort of automated analysis, though.
posted by SpacemanStix at 10:12 AM on June 5, 2015 [1 favorite]

MetaFilter: It'd be far from perfect but it'd be something.
posted by ODiV at 12:24 PM on June 5, 2015 [1 favorite]

With stuff on Ask you could at least take a stab at an automated analysis; look at each question with at least one best answer and compare the timestamp of the question to the timestamp of the first best answer. It'd be far from perfect but it'd be something.

For questions with a best answer recorded, the median time to a best answer is 2190 seconds (36.5 minutes). The mean is 55060 seconds (15.3 hours), a lesson in how outliers can badly skew the mean.

The fastest times are a bit sketchy because daylight saving time weirdness can result in apparent negative times. Throwing out negative times (but otherwise not doing anything to control for this) gives the following top 10 fastest best answers:

postid commentid diff_in_seconds
187326 2696887 20
156215 2239056 20
188864 2718274 21
42191 648933 23
28092 442203 24
82059 1215523 25
162970 2340719 28
134680 1924161 29
137413 1963848 29
206904 2982066 31

(I don't recall how to make a proper table on MeFi.)

There are 2452 best answers with a timestamp less than or equal to 180 seconds after the question's timestamp. There are 105,590 questions with at least one best answer and 275,639 questions total, so the "3 minutes or less" category accounts for 2.3% of all answered questions and 0.9% of all questions.

Going up to 5 minutes gets 6504 answers, or 6.5% of best answered questions and 2.4% of all questions.

If you like histograms, here's some data:

seconds_to_best_answer count pct_of_answered pct_of_all_questions
0 58 0.05 0.02
100 1565 1.48 0.56
200 2995 2.83 1.08
300 3860 3.65 1.40
400 3843 3.63 1.39
500 3802 3.60 1.37
600 3599 3.40 1.30
700 3466 3.28 1.25
800 3131 2.96 1.13
900 3005 2.84 1.09
1000 2706 2.56 0.98
1100 2643 2.50 0.95
1200 2329 2.20 0.84
1300 2203 2.08 0.79
1400 2051 1.94 0.74
1500 1967 1.86 0.71

And here are cumulative values through the first half of all questions with best answers:

seconds_to_best_answer count %_of_answered %_of_all_questions
0 58 0.05 0.02
100 1623 1.53 0.58
200 4560 4.36 1.66
300 6855 8.01 3.06
400 7703 11.64 4.45
500 7645 15.24 5.82
600 7401 18.64 7.12
700 7065 21.92 8.37
800 6597 24.88 9.5
900 6136 27.72 10.59
1000 5711 30.28 11.57
1100 5349 32.78 12.52
1200 4972 34.98 13.36
1300 4532 37.06 14.15
1400 4254 39 14.89
1500 4018 40.86 15.6
1600 3728 42.52 16.23
1700 3397 44.06 16.82
1800 3143 45.48 17.36
1900 3009 46.9 17.9
2000 2846 48.17 18.38
2100 2647 49.4 18.85
2200 2566 50.59 19.3

Caveat: I didn't bother differentiating between deleted and undeleted questions.
posted by jedicus at 1:16 PM on June 5, 2015 [15 favorites]

I should say that all of that data came from the May 30, 2015 infodump.
posted by jedicus at 1:18 PM on June 5, 2015

You're doing Glob's work, jedicus.
posted by cortex (staff) at 1:27 PM on June 5, 2015 [3 favorites]

Maybe. Looking at it, my histogram and cumulative tables don't quite match the ad hoc numbers I calculated for 180 seconds and 300 seconds, so there may be some slop in there.
posted by jedicus at 1:45 PM on June 5, 2015

Wow, that's great. Thanks jedicus! Even if it's not perfect, it provides a pretty nice proof of concept with interesting data in it in its own right.

There are 2452 best answers with a timestamp less than or equal to 180 seconds after the question's timestamp.

I knew we were awesome.
posted by SpacemanStix at 1:55 PM on June 5, 2015 [1 favorite]

Cutting out the relationship-tag questions would probably be more realistic for the time to best answer, for questions which actually have an answer, so to speak.
posted by Rumple at 10:56 PM on June 5, 2015 [1 favorite]

275,639 questions! That seems like a lot.
posted by notyou at 7:49 AM on June 6, 2015

For anybody who reads peagood's link and wonders where "things in these proportions" went, here's a Wayback Machine link to the image.
posted by cgc373 at 2:31 PM on June 6, 2015 [1 favorite]

I bet the stats are skewed who mark every good answer as the best answer. Wonder if you can write a query to limit it to questions with one best answer.
posted by smackfu at 12:08 PM on June 10, 2015