LogFAQs > #966277466

I, too, would like to apologize for jumping to conclusions that anyone here was trying to get Raetsel's ranking this week discounted or anything like that. Clearly there's a bit of a gulf between my beliefs on giving posts the benefit of the doubt due to the difficulty of conveying tone/intention in text and my willingness to actually give said benefit of the doubt. Instead, I let some strongly held feelings I have about the nature of art and engagement with it take over in an attempt to stave off what I feared was happening, and again, I'm sorry for that rather rude misinterpretation.

Uh anyways, now that I know we're all on the same page, here's some (inter)related rambles that you all probably shouldn't read:

~-~
(On Statistics [a reply to #19])
~-~

BlueCrystalTear posted...

I was thinking just extreme outliers being discounted for statistical calculations. Like consider the formula where x = the number of album tracks:
x * ((X-1)/2)

So 14 * 6.5 = 91.

It would be 105 for 15+ tracks, 78 for 13 tracks, 66 for 12 tracks, 55 for 11, 45 for 10 tracks, 36 for 9 tracks, 28 for 8 tracks, and 21 for 7 tracks.

I probably should know the answer to this, but I am curious - what is your reasoning behind using that specific formula to determine whether or not an outlier is extreme? I can see that said formula offers a smooth progression from an average outlier of 3 ranks per track at 7 tracks up to 7 ranks per track at 15 tracks in .5 intervals, but I'm sure there's something more foundational I'm missing. And yes, I'm also sure I'm not living up to my username by this line of questioning >_>

BlueCrystalTear posted...

As someone with "Stats" in your username, I'm sure you know about outliers which skew the average in ways that don't accurately represent consensus.

I do indeed, partly from personal experience - though our track rankings don't directly draw on the average of each participant's individual ranks, which likely dampens the effect somewhat. To say much more directly about what I quoted would be a completely unnecessary rehash of earlier posts. However, I would like to raise the related point that individual song rankings can likewise skew that track's average in unrepresentative ways.

A case in point that I alluded to before is my #1 rank for Fitter Happier when we tackled OK Computer - a drastic outlier with only one other rank above #9, even though my album outlier score was unexceptional. It was only my third time participating in Rank the Tracks, ranking two minutes of quasi-spoken word atop an album that I knew had an exceptional mystique among music fans of all sorts, and I didn't even attempt to specifically justify that ranking. For all CZ would have known back then, it was an act of trolling on my part. Yet they allowed my ranking through anyways, for which I am forever thankful even though I ungratefully whined about something semi-related at the time.

Anyways, this statistical talk inspired me to check the stats (specifically, standard deviation) for how great an outlier a few of my most unusual rankings truly were. The results notably confirm my Fitter Happier ranking as a (seemingly my most, in fact) drastic departure from the norm; I'll post them tomorrow as part of a post I haven't quite finished tonight.