Visualization Implementation (VI5)

README.md

John Berlin CS725

Visualization Implementation (VI5)

Rule of Thumb

I choose to implement the rule of thumb: No Unjustified 3D: Occlusion Hides Information

I used R to achieve this as I have been working a lot in it. The plot that breaks the rule of thumb is seen here

The rule of thumb is necessary because:

1. There are so many circles that we can not tell in a glance at what position they are or how many at the position.
2. Axis labels are hard to understand especially how the z-axis correlates with the other axises.
3. Even tho the color red indicates depth it does not help when there are more dots in the forefront than allow for the background dots to be seen
4. Still color indicating 3d depth is lost at the plots angle.

This rule of thumb is necessary to prevent issues as briefly mentioned above from happening

Broke Rule of thumb code

library(scatterplot3d)

d <- read.csv("tileDif2.csv",stringsAsFactors = TRUE)

# The xaxis is the number of time alsum won
# The yaxis is the number of times another method won
# The zaxis is the histogram differences 
scatterplot3d(d$awon,d$mwom,d$histDif,highlight.3d=TRUE, 
              angle = 25, scale.y = .5,
              xlab="Times Alsum Won",
              ylab="Times Another Method Won",
              zlab ="Histogram Differences")

Fixes Rule of thumb code

In order to fix that the 3d plot a regular scatter plot is used and jitter(an over-plotting correction method) is applied to the points. Also color indicates the difference metric. The color legend was omitted due to it being larger than the plot itself.

Rather than having to deal with figuring out where in relation to 3d space each dot is the jitter provides us the same effect but with better detail.

One could argue the jitter applied makes it seem like there are values such as 1.5 due to the dots being in between two numbers.

But it is simply better than having to scratch our heads about where in 3d space is each point. Yes even though the color scheme is very bad it more clearly shows to us the difference attribute than does 3d.

library(ggplot2)

d <- read.csv("tileDif2.csv",stringsAsFactors = TRUE)

#To better express the same 3d plot in 2d we add jitter to the scatter plot to avoid overplotting
ggplot(d,aes(x=awon,y=mwom))+
  geom_point(shape=1,position=position_jitter(width=1,height=1.5),aes(color=factor(histDif)))+
  scale_x_continuous(breaks = round(seq(min(d$awon), max(d$awon), by = 1.0),1)) +
  scale_y_continuous(breaks = round(seq(min(d$mwom), max(d$mwom), by = 1.0),1)) + 
  labs(x="Times Alsum Won",y="Times Another Method Won",color="Histogram Differnces")

RuleOfThumb.R

library(ggplot2)
library(scatterplot3d)

d <- read.csv("tileDif2.csv",stringsAsFactors = TRUE)

scatterplot3d(d$awon,d$mwom,d$histDif,highlight.3d=TRUE, angle = 25, 
              scale.y = .5,xlab="Times Alsum Won",ylab="Times Another Method Won",zlab ="Histogram Differences")

ggplot(d,aes(x=awon,y=mwom))+
  geom_point(shape=1,position=position_jitter(width=1,height=1.5),aes(color=factor(histDif)))+
  scale_x_continuous(breaks = round(seq(min(d$awon), max(d$awon), by = 1.0),1)) +
  scale_y_continuous(breaks = round(seq(min(d$mwom), max(d$mwom), by = 1.0),1)) + 
  labs(x="Times Alsum Won",y="Times Another Method Won",color="Histogram Differnces") + 
  guides(color=FALSE)

tileDif2.csv