Graphing Your Data: Graphing Software and Other Tips

1
Graphing Your Data Graphing Software and Other
Tips

• Priti Shah

2
Goals

• Communicating an important quantitative fact
  (e.g. an interaction) to an audience (talk,
  journal article)
• Exploratory Data analysis (trying to figure out
  what your data say

3
Software options

• Macintosh Software
• Cricketgraph
• Deltagraph (my personal favorite)
• Microsoft graph
• PC (I dont know)-- oh they have deltagraph!!!
• Exploratory data analysis
• Most statistic packages have some graphing
  capabilities (boxplots, stem-and-leaf plots,
  scatterplots, etc.) S is terrific.
• High-end software (Inxight, Lucent)

4
Most have similar structure

• Table(s) of data
• Choice of graph type
• Click on columns you want to graph
• Very simple!!

5
Deltagraph

• Made by SPSS
• http//www.spss.com/software/deltagraph/
• Demo available

6
You can make beautiful graphs (from Deltagraph
Web site)
7
Another Sample
8
Another one
9
More samples
10
Delta Graph What makes it special?

• Formats availableArea,Bar,Bubble, Column,
  Contour, Pictograph, Pie,Scatter,
  Time Line,
  Surface,Wireframe,Histogram, Box Plot, Bullet,
  Organization Table.
• Easy to use
• Can make it fancy

11
Using Delta Graph

• Open Delta Graph
• An Untlitled File will open save as something
• A data page will open. There are data pages
  (pages of tables) and chart pages that are
  pages with graphs on them.
• Enter data (going downwards will be x axis,
  across will be z if you have 2-variable design)

12
Consider, for example

• Study on Perceived values of numbers (with
  Johanna)
• Asked subjects to judge how good (on likert
  scale) different proportions would be, in
  specific contexts (a grade of 8/10, a grade of
  80/100).
• Variables
• proportion (.05-.95)
• Scale (10, 100, 1000)
• Context (grades, toothpaste, cancer, profits)

13
Entering data

• We want to plot perceived goodness as a function
  of proportion and scale for each context (4
  different graphs)
• First thing we do is enter our data (this is like
  a cooking show). We have different tables for
  each context (Data pages). Use table icon to go
  to new data page. Name each page so you know
  what context it is. (to name page, click on
  view on menu then go to name page)

14
Making a graph

• Highlight data you want to graph
• Click on data on menu page and go to Chart
  Gallery
• Choose a chart (line graph)
• A graph will automatically be created with
  certain defaults on a new chart page
• Name the chart page by going to view and name
  page

15
Changing features of our graph

• Click on text box to add text
• Using the chart menu
• Go to axes to change scale or start point, etc.
• Use labels to label x and y axes
• Use ticks and grids to change where your ticks
  are, and whether or not you have grids
• Use toolbox on left to change colors, patterns,
  add text, draw anything, etc.
• Change size by pulling on corner move

16
Making a new graph exactly the same

• Go to file menu to open up a new library
• Drag your graph into the library and name it (you
  can either save as template or save w/ data).
  Save as template
• Now, drag the template back to your chart page
  (if you want it on the same page). Or, open new
  chart page by clicking on graph icon and new
  page
• It will now go to the original data page.

17
Making new graph continued

• Now, go to a new data page (lets look at breast
  cancer page). Highlight the data you want to
  graph.
• Click on plot.
• Now a graph with same size and features appears
  on the page (you may have to make some changes,
  such as scale or labels, depending on data)

18
Tips for creating good graphs

• Communicate relevant information explicitly
  (think What do I want to convey). Remember,
  different graphic formats highlight different
  information about the same complex data. Consider
  the following examples

19
Look at this graph
20
Compared to this one
21
Or this one
22
Compare this graph
23
To these
24
Specific Tips

• Use line graphs for conveying trends (graphic
  convention is to only use for continuous data,
  but I violate that to make my point sometimes)
• Bar graphs to make categorical comparisons
• 3-d graphs if exact data isnt important, but a
  complex relationship or some idiosyncratic data
  points are important.
• Pie charts are not good for exact values, good
  for comparing proportions.

25
Avoid forcing the reader to make any
calculations or complex inferences

• Do not make them mentally change scale
• Do not make them subtract everything from 100
  (e.g. if youre talking about error data in the
  graph and accuracy data in your paper)
• Do not make them subtract values on one graph
  from another graph on a different page!!

26
Keep it simple

• Remember, adding 3d is not always so good. You
  lose ability to read exact data points, and you
  can occlude important data.
• Lots of colors and designs are cool, but not
  necessarily helpful
• Label your lines, avoid using legends
• Make your symbols or lines distinct if they are
  to be compared. Esp if data are complex!

27
And plot only what you want to show

• If the 1000s scale doesnt matter, dont plot it
• If the important thing is the relative values of
  different contexts, plot those four averages in a
  single bar graph. Leave out the other variables
• Etc.

28
Consider

• Goal of graph reader
• Expertise of the graph reader

29
Consider some bad graphs

• Like this 3d graph of our proportion judgment
  study

30
Using graphs for more

• Exploratory data analysis
• Many statistical software packages allow you to
  make simple graphs to look at your data.
• Some quite sophisticated
• Consider some of these high-end tools

31
High end softwareVisualinsights Advizor
32
Inxight (xerox parc)