Log in

No account? Create an account

fuzzy week

Last night I went to a party with Ellen where there was confetti all over the floor and streamers and used car lot style shiny patriotic triangles on the walls and drank some whiskey out of the bottle. There was a pile of free things and I rescued a Cadbury branded clucking rabbit from it. At some point, the poppable packing materials found their way onto the carpet and much popping of them ensued. Earlier, there was a small white dog that was very excited about a piece of rope.


Today my biggest accomplishment has been listening to the KEXP top 90.3 of 2005 [kexp (starts at 10 am, 30 december 2005)]. [Except that right now they're playing "Kissing the Lipless" even though Chutes Too Narrow came out at the end of 2003. Ah, they're cheating by including their live compilation disk]. I'm so glad that I took evan's suggestion [#] about downloading the streaming mp3 file because I'm thoroughly out of practice in the "listening to the radio" department and if I couldn't fast-forward past the surprisingly long surprisingly inane breaks of DJ chatter, I think that the experience of not being able to just pick the songs that I want to listen to in the order I want to listen to them might just drive me completely crazy. I couldn't find an mp3 player that completely understood that the track it's playing is getting larger with time so I occasionally need to reopen the file. The point of this is to say that carrying around a bunch of music that I like has perhaps spoiled me and over-indulged my control freak work environment tendencies.

I thought that my big accomplishment would be sending some FAXes, but that has been a bust. Alas, there's still time to conquer the long-distance dialing instructions.

updates: as the playlist gets into the 20s I'm liking it more and realizing that the most stressful part of radio is the people talking directly into my ears (which may be a small part of my telephonehate). Also, Mission FAX Machine has come to a successful conclusion.


Yeah, I noticed a few other "cheats", e.g. Postal Service put out a B-side or something this year, so they played a track off it but that track also happened to be on Give Up.
apparently listeners voted on these? I wonder if it was free-response or if there were a fixed set of nominations.
There were 5 dropdowns with everything on their playlist from 2005, and a text-entry box for write-ins. Options in the dropdowns were alphebetized, so I wouldn't be surprised if the results were skewed towards the beginning of the alphabet.

In fact, notice how nobody in the top 6 (other than Sufjan, who won by "a landslide" anyway) is lower than the letter "D"...

BTW, the other "cheat" I forgot about was The Commander Thinks Aloud, which came out in 2004 on the MFA(?) comp, but the Long Winters re-released it on an EP this year.
were the 5 dropdowns to pick your top 5 songs or to split up the alphabet?
IIRC, they were the same 5 dropdowns, one for each of the top 5.
It doesn't look especially biased by first character of track title or artist name:

You should put bigger circles around higher ranked artists, and see what happens then...

BTW, you are a graph wizard.
I'm curious -- were the things in the drop down lists sorted by track name or artist name?

Neither is statistically significant, but the regression coefficient for first title of track is -0.38 (95% confidence interval, -1.2 to 0.48), suggesting that each unit increase in first character is associated with a 0.38 lower ranking. Except that "lower" means closer to zero.
Sorted by artist name. Since it's the top 90.3 Albums, people were voting on an Artist - Album pair.

I can't remember if artists whose artist name is a proper name (e.g. "Laura Veirs") were sorted by first name or last name. Nor can I remember if artists starting with an article (e.g. "The Decemberists") were sorted by the first non-article letter, or the first letter including the article.
Also, the number of artists available for each letter of the alphabet should be taken into account, so if 5 out of 5 A's were in the top 90, but only 15 of the 50 "Wolf" bands, a normal histogram wouldn't show an alphabetic bias.
I have no idea how to figure that out, but here's the distribution of first letters of artist names:

c | Freq. Percent Cum.
( | 1 1.10 1.10
A | 6 6.59 7.69
B | 15 16.48 24.18
C | 7 7.69 31.87
D | 5 5.49 37.36
E | 1 1.10 38.46
F | 3 3.30 41.76
G | 2 2.20 43.96
H | 3 3.30 47.25
I | 2 2.20 49.45
J | 1 1.10 50.55
K | 4 4.40 54.95
L | 5 5.49 60.44
M | 9 9.89 70.33
N | 3 3.30 73.63
O | 2 2.20 75.82
P | 2 2.20 78.02
Q | 1 1.10 79.12
R | 2 2.20 81.32
S | 9 9.89 91.21
T | 5 5.49 96.70
W | 3 3.30 100.00
Total | 91 100.00

Does this look like they're similar to the distribution of words in the english language? even if not, does it look similar to the distribution of band names?
Really wish I would have saved the voting page, I can't recall the distribution of band names. Perhaps emailing KEXP would help?

I'm not sure if it's safe to plot band names against the distribution of words in the English language, since many artists are proper nouns, foreign words, nonsense words, and general non-vernacular (who would go around saying "Helsinki" if it wasn't for AiH?)
Without getting the list of choices (who to even e-mail? and it's not in archive.org), the best evidence for your theory may be the "B" band names (articles removed, not sorted by last name): B is the most common first letter among winners, while my google searching suggests that B is the eighth in Webster's [google]. Not very authoritative, I know.

In any case, the contents of list itself had to have biased the results more than the order.
I just filled out KEXP's "support request" form, asking for a copy of the voting page. Let's see if they are nice.

You're right about the dropdown itself being a strong form of bias.
kexp got back to me, here's the voting page:


"The" bands are listed under "T", "(Various)" compilations are listed first, proper name artists are listed "First Last"
interesting. I'll bet that that order helped the KEXP comp's placement on the chart!

I'll look at this more tonight to see if the odds of being on the chart are related to place in the alphabet.
cool, thanks


breaking news! each increase in letter of the alphabet (e.g. having a band name starting with a "b" instead of an "a") in the list of choices was associated with a 3% lower chance of ending up in the top 90.3.

the odds ratio for each unit change in first character was 0.97 (95% confidence interval, 0.94 to 0.9998).

I guess metroblogging seattle chose its favorite scandal of 2005 too soon.

Re: developing...

should we bring this scandal to the attention of the media and/or KEXP?

Re: developing...

I should probably think about it a little more. The p-value for the trend was 0.0479, which is just on the inside of conventional levels of statistical significance.

I don't know about issuing a press release, but maybe I could turn it into a metblogs post?

Re: developing...

I think it would be curtious to notify KEXP first.