New Wave Outpost Forum

by **Frau_Blucher** » Sat Dec 10, 2005 5:30 am

Okay, this next chart is so freaking cool I wanna scream like a little girl...but I won't...or maybe I will. [:I]

Taking some suggestions, I designated the Top 200 bands (I couldn't deal with all the onesie-twosie crap) as more Guitar/Red or Synth/Green ones with Red in positive numbers and Green in negatives. One class is more punkish and one more synthish. Obviously there's lots of overlap. Some heavy synthy stuff I put as Red (like Bauhaus) while some light guitar stuff I put as Green (like Talk Talk). It's a quick and dirty test, the results will hold up on the whole, don't nobody fooking argue with me! [:D]

I ran this test using Complete Correlation wih Euclidean Distance Clustering since these numbers are more real, having both Magnitude and Direction rather than just a Rank Order. I listed some sample Correlation Coefficients but not all of them - you can compare your's by just approximating across the Tree. We still see LovePlusOne the most tightly correlated with the Average - good, that validates we didn't monkey with it completely. But lots of other ones have changed and I think tellingly!

The really big thing I see is the two distinct bins on the Left and Right. My interpretations:

Large Left Bin: this looks like your typical mixed population of both punkish and synthish new wave, including the Average. I'll buy that.

Left Bin Left versus Left Bin Right: that Left Bin breaks down into two distinct bins. Scanning down the colors, I'd say the left one has a bit more Guitar mix while the right one has a bit more Synth mix. I totally buy that for myself and the others sitting in that tree. Hey Manekineko, you feelin' me?! The right one with 2Nu and MikeP has a bit more synth sprinkle - again, dead on IMHO. And I'm sure Marv is much chagrined that, horrors, he's so Average! [}:)]

Right Bin: this is where things get really interesting. Everyone isn't similar in this bin internally, just different from the other bin. I see three strong, unique nodes within this bin.

Right Bin Right: this is definately the Synthy or Poppy bin. With Schwenko sitting in there, I totally buy this result. BTW, do you see Sean WAY out there, completely detached from everyone else? He's so exclusively synthy, he's not correlated with ANYONE else, AT ALL! Wow.

Right Bin Middle: this one is a tough call because it looks like a mixed bin like the Far Left one. It's just that its mix of the mixing is different than the standard mix. Kudos to your well-rounded yet still different tastes!

Right Bin Left: this one is perhaps the most fascinating yet. Definately the Punkish bin. There's two branches though. The Right more Punkish branch and maybe the Left more guitar/alt branch. Looking at Motors, Rubbeet, and Waukena in the right one, I think no one would argue! How freaking cool! I think I hear wedding bells...Wauk and Motors will most assuredly live happily ever after!!!

<img src="http://www.pbase.com/myirwin/image/53382742/original.jpg">

by **manekineko** » Sat Dec 10, 2005 5:54 am

So let me make sure I understand. The left bin is people who like a mix of guitar and synth. The right bin is people who have a preference for one over the other, either guitar or synth. Is that correct? I am a little confused because I would have thought the data would be displayed as a range with guitar on one side of the graph, synth on the other side, and a mixture in the middle.

[quote][i]Originally posted by Blir[/i]
<br>Okay, this next chart is so freaking cool I wanna scream like a little girl...but I won't...or maybe I will. [:I][/quote]

Go ahead, scream like a little girl! The chart is so freaking cool!

[quote]Hey Manekineko, you feelin' me?! [/quote]

Yeah, and it's the biggest thrill I've had all day! [:p]

by **Frau_Blucher** » Sat Dec 10, 2005 6:01 am

[quote][i]Originally posted by manekineko[/i]
<br>So let me make sure I understand. The left bin is people who like a mix of guitar and synth. The right bin is people who have a preference for one over the other, either guitar or synth. Is that correct? I am a little confused because I would have thought the data would be displayed as a range with guitar on one side of the graph, synth on the other side, and a mixture in the middle.
[/quote]
Mostly right Manekineko. But this data doesn't occupy a continuous spectrum from one side to another another. If you think about it, it first starts as binomial negative or positive data - Greens or Reds. Then we add magnitudes to it with our rankings and further mix it up by choosing both Reds and Greens in a highly heterogenous manner from person to person. That's why it falls into bins rather than a continous spectrum - there's too many tangents off a single stream, "multivariate".

You're right about the Left bin and mostly right about the Right bin. But that Right bin also has a mixed bin that are just mixed differently than the Left bin.

by **NuWaveRx** » Sat Dec 10, 2005 6:19 am

Which puts me about where I thought I be for such an analysis-- mostly synth with some guitar-oriented bands in there. Hey Blir was my assumption correct about the strength of correlation?

by **2Nu** » Sat Dec 10, 2005 6:23 am

Musically Inbred?! Moi? Hrrumph.. Well it's gotta be the water.. or perhaps, a smidgen of WLIR subliminal influence.

Great Job Blir![;)]

Nice choice of holiday themed charting colors BTW.

<IMG SRC="http://img.photobucket.com/albums/v395/richp62/Icons%20Avatars/inot.gif">

by **manekineko** » Sat Dec 10, 2005 6:24 am

So how did those mixed people end up in the right bin? What was it that put them in the right bin instead of the left? I'm hoping there is a simple answer to this. [:o)]

by **Frau_Blucher** » Sat Dec 10, 2005 6:24 am

[quote][i]Originally posted by NuWaveRx[/i]
<br>Which puts me about where I thought I be for such an analysis-- mostly synth with some guitar-oriented bands in there. Hey Blir was my assumption correct about the strength of correlation?
[/quote]
Oh, right, the close the number to 1, the stronger the correlation...based on the inputs of course! You know the bit about the 3 kinds of lies: Lies, Damn Lies, and Statistics!

by **Frau_Blucher** » Sat Dec 10, 2005 6:57 am

[quote][i]Originally posted by manekineko[/i]
<br>So how did those mixed people end up in the right bin? What was it that put them in the right bin instead of the left? I'm hoping there is a simple answer to this. [:o)]
[/quote]
LOL...software! Okay, seriously? Short answer is "NO", there is no simple answer. The long answer? Oh boy, I'm not even sure...

Okay first, we've got things (bands) being measured by us (ranks), right? Between any persons A and B, there will be a difference or distance between those two measurements, right? This distance between every band amongst and between every person needs to first be calculated - this would be staggering without a computer!

For the simple similarity comparison, I used Pierson Rank correlation, which is very useful when you have large outliers - it doesn't take into effect the magnitiude for example of Band X, just its ranking. It's a good true measure of broad similarities.

For this latest graphic however, I used a Euclidean Distance correlation since we had things that certainly had Rank but also Magnitudes that were very important now, more than just Rank, because they also could go in opposite directions.

When all these individual components have been correlated, you get this "Distance Matrix" from which all the elements as a whole need to be Clustered based on similarity. I used a flavor of Heirarchical Clustering, all of which basically do the following:

First step was getting that Matrix. Then clustering begins - an agglomerative heirarchical clustering process where repeated cycles of the two closest remaining items (Outpost Members) are joined as branches on a tree. That pair is then replaced by an item that represents that new branc and the calculation is repeated to get the next closest item and so on until you have none left. Each branch will then have a correlation measurement based on various types of distance calculations - the magnitude of that measurement is only relative and not comparable to different Correlation methods, which is why the first graphic has smaller Correlation numbers than the last one but they're not cross-comparable.

by **KYYX4ever** » Sat Dec 10, 2005 9:12 am

Holy synth, that is so geeky and beautiful, I love you Blir! [:I]

I actually don't understand most of this, but the charts and colors sure are pretty, and I love seeing how similar and dissimilar we all are for certain bands.

by **abacab66** » Sat Dec 10, 2005 9:40 am

I think I'm going to need an operation to get myself unjoined at the hip from 2NU.[;)] Actually I'm proud to be in such upstanding company. Thanks for the hard work Blir!!!!!

by **BigBri** » Sat Dec 10, 2005 12:06 pm

Thanks, Blir. This is really cool!

I'd like to say something really profound, but since I don't understand a thing here (surely a product of inbreeding-- I do live on an island, after all), I'm going to just smile and nod politely.

[:D] [:D] [:D]

by **schwenko** » Sat Dec 10, 2005 12:24 pm

How could I have left out Simple Minds from my list???
I demand a recount!![:D][:D][:D][:D]

by **Spit** » Sat Dec 10, 2005 1:52 pm

Blir, It looks like you put a lot of time and work into this but I'm still not quite sure who the top 25 bands are . [:p]

by **My Aural Stimulator** » Sat Dec 10, 2005 2:39 pm

Blir, this is totally amazing. Totally.

by **Rubbeet921** » Sun Dec 11, 2005 3:42 am

Holy cow!!! You never fail to amaze me, myblirwin! This definitely was a huge labour of love! What will we ever do without you? [:p][:D]

New Wave Outpost Forum

Ultimate NWO Band Poll - More Results!

Who is online