Forums  > Software  > How to draw the yield curves in time and paths?  
     
Page 1 of 1
Display using:  

pj


Total Posts: 3318
Joined: Jun 2004
 
Posted: 2017-01-19 16:54
Let us imagine I have a number of paths (say 100 )
with 10 steps each.
On each step of each path I have a yield curve
generated.

I mean, finally I have 4 dimensions.
time, path number, and the yield curve which is two dimensional.

How to present that visually?
if I had only a point on each step of each path
I could draw a beautifully coloured surface. Like this



But how to do that with curves?
More over all them are going from zero to 1.
I am sure that question was already tackled,
but my Google skills are not up to the task.

Any ideas?

The older I grow, the more I distrust the familiar doctrine that age brings wisdom Henry L. Mencken

tabris


Total Posts: 1244
Joined: Feb 2005
 
Posted: 2017-01-19 17:14
hmm whats wrong with interpolation of the 10 steps?

Dilbert: Why does it seem as though I am the only honest guy on earth? Dogbert: Your type tends not to reproduce.

pj


Total Posts: 3318
Joined: Jun 2004
 
Posted: 2017-01-19 17:20
I am not sure what you want to say.
Should I show the average of yield curves on each path?
Well the yield curves are different on each step so I will be
losing quite lot of information.

The older I grow, the more I distrust the familiar doctrine that age brings wisdom Henry L. Mencken

Maggette


Total Posts: 943
Joined: Jun 2007
 
Posted: 2017-01-19 17:42
Well, I know sub optimal, but is an animation an option for you?

Like a moving surface (over one dimension)?

Ich kam hierher und sah dich und deine Leute lächeln, und sagte mir: Maggette, scheiss auf den small talk, lass lieber deine Fäuste sprechen...

pj


Total Posts: 3318
Joined: Jun 2004
 
Posted: 2017-01-19 18:10
> Like a moving surface (over one dimension)?
I was considering it, but maybe something nicer?

The older I grow, the more I distrust the familiar doctrine that age brings wisdom Henry L. Mencken

svisstack


Total Posts: 300
Joined: Feb 2014
 
Posted: 2017-01-19 19:25
slider for adjusting step_number/time under the graph

Time well wasted.

pj


Total Posts: 3318
Joined: Jun 2004
 
Posted: 2017-01-19 20:14

Sliders?
Hmmm....
But you won't see the totality.
For the prices you would see the risk zones,
and ask what causes them.


The older I grow, the more I distrust the familiar doctrine that age brings wisdom Henry L. Mencken

Tradenator


Total Posts: 1579
Joined: Sep 2006
 
Posted: 2017-01-19 21:23
Maybe try this or this for inspiration?

pj


Total Posts: 3318
Joined: Jun 2004
 
Posted: 2017-01-20 09:33
Nice movie!
As for inspiration... Sad

The older I grow, the more I distrust the familiar doctrine that age brings wisdom Henry L. Mencken

Tradenator


Total Posts: 1579
Joined: Sep 2006
 
Posted: 2017-01-20 10:42
I think you are looking for a means to show a tesseract to the novice. Projecting into 3D is one way to think of it (eg a movie). I don't know a good answer for how to do this elegantly, sorry. If you want to see how not to do it, there is a Carl Sagan thing on you tube somewhere.

pj


Total Posts: 3318
Joined: Jun 2004
 
Posted: 2017-01-20 10:46
what has a teseract to do with yield curves? Confused

I thought that my original question was straightforward enough.

The older I grow, the more I distrust the familiar doctrine that age brings wisdom Henry L. Mencken

Tradenator


Total Posts: 1579
Joined: Sep 2006
 
Posted: 2017-01-20 10:54
Tesseract = 4 dimensions, just like you have. I thought it might be a simple geometric analogue to your yield curve visualization problem. Sorry if this isn't helpful, but good luck with it.

deeds


Total Posts: 346
Joined: Dec 2008
 
Posted: 2017-01-21 11:43
Seems a question of which information to lose (all models, blah blah blah, all visualizations, ditto)...

your question may come up because there hasn't been enough brute force survey of 'raw data' (curves)...or there is fear that data set in hand is not comprehensive and you want to do raw data survey with your visualization in an online way.

Seems the question about how to show will be easier to answer when you know what features are essential, invariant to your purpose, what information is not relevant.

Designing a tool to make the 'raw data' survey easier is different than designing a visualization which presents the data in most significant cases...

I wouldn't discount the effectiveness of a set of indexed sheets, one per curve which can be arranged and rearranged in meat space.

then consideration of projection (geom/stat), embedding, crypt recoding, measure, summary, interactivity?

Rashomon


Total Posts: 166
Joined: Mar 2011
 
Posted: 2017-01-26 20:50
pj, without more information about what you want to show I think it's hard to give general advice other than the obvious mathematical options. If you want to show certain risk zones you should color them something obvious, like red. Do use Lab colors rather than HSV if you want to code information into the color parameter, but generally I think people lean too heavily on the viewer's ability to pick up precise differences in color/time variation in a chart. This is what's wrong with Chernoff-Fleury faces, and with the FMS Symphony: yes, you can code high-dimensional data onto these parameters, but it just doesn't hit the viewer between the eyes the way an Economist chart does. The eye can detect small quantitative differences in lines/curves, but not #FF0000 − #EE0000 = #009900− #008800. (Nor rotations thereof, nor doing the same with eg a saturation dimension that keeps lightness/brightness/value constant).


It's all audience dependent (imagine below the audience remembers each month/year personally), but I think an animation can look quite good:



And better than coding time into color:




(The above makes the extra mistake of using the whole hue wheel -- because it's there! -- when a difference in opacity would have sufficed, and maybe a bit of variation in hue if they wanted to make it pretty as well as clear.)

The time version of the above plot allows for another improvement, namely doubly presenting the time information. (Which is less obvious than curve fluctuations --- so mentioning that it's 1 month per second as well as writing the changes in months is good.)





Now, let's say you want to draw attention to the 8th month. Then I would code using all greys of different lightness (whether long or near gets more transparency depends on your perspective) and red the 8th curve.

You could also freeze that curve across all time-points, or maybe opacity=1% all the curves and have opacity become 80%-90% as time ticks through that particular curve.



It's very easy to get lost in the computer defaults or in the fun of thinking about high-dimensional math. But attention and clarity is what you're after, and those have mathematical properties that are totally unrelated to whether in theory it's possible to make some bijection between (or epi onto) colour/time/a plane and (from) N dimensions.





I like the above image better than the surface you showed because it has the BAM factor. Evolutionarily our red cones are there to pick out fruit from among a see of green leaves. Your surface picture treats colors like wavelengths, which is physically right but psychologically wrong.





I said greys + 1 red, but you could reinterpret that with browns + 1 blue + 1 yellow (brown is neutral, both blue and yellow can contrast) if you needed to indicate two different things.


I would also suggest that generating everything programatically is inferior to printing out the fine / precise parts and then drawing with a crayon / marker / MS Paint over the parts of the curve/surface you want to highlight. Let's say you want to highlight a small segment of one of the curves. Nobody who wrote matplotlib or R plot expected you to do that, and you don't want to go hack on the libraries / language just to do something so simple. Those intuitions you would do if a crayon / marker were your presentation tool are, to me, more likely to be the right ones. The computer comes in when absolute precision is needed (eg, the hand is bad at drawing slopes that are twice as steep, let alone 10% or 5% steeper). Carl Dair, iirc, said that type is a perfect form of handwriting---a hand that forms every letter and spacing around it exactly the same each time, the letters well proportioned to each other (depending on the font), etc. That (along with the evolution stories about our cones) is what I think about to not get lost in the weeds of my toolkit in presenting something like this.


HTH.



** Incidentally I have a rant stored up about trading screens using green and red instead of warm and cool. Shows very little design / critical thought went into coloring the Big Board.

tbretagn


Total Posts: 245
Joined: Oct 2004
 
Posted: 2017-02-06 13:24
Pj maybe you can try this:
Hypertools

Et meme si ce n'est pas vrai, il faut croire en l'histoire ancienne

pj


Total Posts: 3318
Joined: Jun 2004
 
Posted: 2017-02-06 14:07
Thank you, tbretagn!
That looks quite amazing.
Feedback will follow.
pj

The older I grow, the more I distrust the familiar doctrine that age brings wisdom Henry L. Mencken

deeds


Total Posts: 346
Joined: Dec 2008
 
Posted: 2017-03-10 16:12

sweet.

EGH


Total Posts: 43
Joined: Nov 2014
 
Posted: 2017-06-25 00:08
Interesting stuff. But not exactly Rocket Science! And I should know!

deeds


Total Posts: 346
Joined: Dec 2008
 
Posted: 2017-06-26 18:48

beats thinkin'.

Nelson Siegel Svennson?

Tradenator


Total Posts: 1579
Joined: Sep 2006
 
Posted: 2017-06-27 09:52
EGH, when did you last design a rocket?

EGH


Total Posts: 43
Joined: Nov 2014
 
Posted: 2017-06-27 21:15
"EGH, when did you last design a rocket?"

Last night actually! Why design the rocket, when you can simply mix the rocket fuel: rocket
Previous Thread :: Next Thread 
Page 1 of 1