Code here
Executable here

Conclusions

 I really like the ability to get quick patterns or feel of the data, that parallel coordinates (P.C.) give. Also they let you see a bunch of variables at once. This is very hard to do any other way...in normal 3D views...you can at most see 5 dimensions if you use shape and color but even then, its gets very confusing very quickly.
Since P.C. graphs can give a general feel of things it is nice to be able to use 3D/4D views to get a better view of patterns that might arise.
For this, it is good to have Delaunay function (which basically makes a nice surface from the points). However this only works well when you the data you are viewing has many points, if your filter is reducing the view data to a small number or low density points, then it is better to just use a plain old 3D graph.
Finally splatting is a lot like the 3D graph though it has the added advantage of blurring the data almost so it is easier to see some larger patterns.
Post finally...I was very pleased with the speed that the computer was able to run and render nearly a million points.

VTK...just the way I like it

I used C# with the .NET dll's yet again...for the obvious reasons...because I learned how to use it from the last two projects and I found that it work well with my object-oriented thought process.

Parallel Coordinates

 I knew I wanted to use parallel coordinates because I really liked the ease in which it reprsented the data, as seen in the car example. So I had to look for datasets that had a lot of variables, which is what I did and I was able to find census data. To implement the P.C. I actually used to vtkParallelCoordinateActors because I needed one as the base and the other as the highlighting. This made things a bit tricky because I had to constantly make sure that the two lined up properly by adding two dummy points that have the same range as the base data into the highlight data.

Why the 3D stuff?

 I added in the 3D plots for many reasons, primarily because I wanted to know what the data would look like in 3D with various dimensions in those axes. As I implemented this I realized that the data does look pretty good and can be helpful in two ways, to help the more 3D oriented people see the data and to help get a more detailed view than the P.C. graph.

Features of the visualization...
Interesting Things
The following table will show you some of my findings...and trust me there are a lot more, if I had more time I could have found more for sure.
So...since most of us are post graduate students, I though I would tell you the best places to work...at least in 1994/95.
The best money is in the private sector...basically working for MicroSoft or something.
But if you can't get in there try either the local or state governments...and then federal.
However, if you are good enough you can be self employed and make a heck of a lot in capital gains, so maybe that is your route.
Though don't expect the really big bugs unless you are lucky :). Also the data shows that a lot of post graduates played the stock market and made money in 1994 and 95.
  • Both years.
  • All data post graduates.
  • With various employment sectors highlighted.

Private Sector

Local Government
State Government

Federal Government
Self Employed
Top view Side view
Clear pattern in the age/weeks worked and occupation type. It is not just random (which I did not expect).
This means that the occupation number must have some sort of pattern to them.
Also the side view shows that age, weeks worked and occupation number don't have a large impact on wage, except for the very old ages...around retirment time, which is to be expected.
  • All Data from 1995
  • Axes = Wage, Weeks, Age
  • Color = Occupation
Same patterns as in 1995 for the most part, just the highest wage was in a different position causing the spike to be in a different place.
  • All Data from 1994
  • Axes = Wage, Weeks, Age
  • Color = Occupation
White Females Black Females
Black females overall have lower wages than white females.
More details:
25% of white females have a college degree of some sort.
Less than 1% of white females have a post graduate degree.
16% of black females have a college degree of some sort.
3% of black females have a post graduate degree.
  • All female data.
  • Highlighting either white or black females respectively.
Married Widowed
Never Married Seperated or Divorced
Married women who are still with their spouses actually make higher wages overall than the other groups. Never married women also make slightly higher overall wages than the other seperated/divorced and widowed. Widowed make the lowest salaries. Some of these results are expected where as others are not, I expected married women to actually be making less money than never married...but...the data says otherwise.
  • All female data.
  • Highlighting various marital statuses.
The wage is mostly in the lower amounts, though they seem to be getting a good amount of stock dividends. Also their capital losses are lower. Another peculiar pattern is that their occupations seem to be clustered in the lower numbers.
  • All female data.
  • Highlighting post graduate female data.
3D Graph Splat
Men have higher capital gains for the higher occupation numbers...as seen by the clustering on the top right of reds. Not exactly sure what this means, but there is definitely a pattern. Also these show the difference between 3D Graph and Splat views (graph is finer than splat).
  • Both years.
  • Axes = Occupation, Gains, Zero Axis
  • Highlighting female data.
Married Never Married
Seperated or Divorced Widowed
There are a lot of patterens that are clearly visible here.
First off, it shows a clear age difference in the higher wage people who are married and those who have never married. This could mean that either people who make a lot of money get married at an older age or that they don't live as long...
Another clear pattern is the age in the widowed individuals...as expected they are older and their wage range though high is not among the top most...and they don't have as many stock dividends...maybe they don't like to invest in the future?
Again as expected...those who are divorced or separated are right in the middle aged group. I am guessing they either die off or more likely get remarried later in life.
  • Both years.
  • Only high wages.
  • Highlighting various data.
No real pattern in capital gains vs. wage, weeks worked, or age. Kind of unexpected because I was thinking that atleast age or wage would have an effect on this. Found this a bit odd...
Also this demonstrates a good use of the 3D Graph.
  • Both years.
  • Axes: Wage, Weeks worked, Age
  • Coloring by capital gains.
48.1% of indian males over 21 are post graduates! (however they don't make much money :(...in 1994/95).
Though my guess is that its because they didn't fill it in...I hope.
  • Both years.
  • Indian males over 21.
  • Post graduates highlighted.
Male post graduates make money but are not in the highest levels...lots of stock dividends though!
  • Both years.
  • Males.
  • Post graduates highlighted.
Ages 21 through 25 All Ages
Females in the range of 21 and 25 years of age make less money then males, but this pattern does not continue through out life.
  • Both years.
  • Various age ranges.
  • Females highlighted.
White people are the most active old people...granted there aren't as many of the other races, but white people's occupations are far more dispersed. Asians and Blacks do have some occupations.
Also none of the other races have any stock dividends...which I find pretty peculiar.
  • Both years.
  • All data for old people.
  • With various races highlighted.

White

Other
American Indian

Asian / Pacific Islander
Black