Skip to content

Quick question: Does this graph have a name?

7 messages · mika03, Duncan Murdoch, Peter Jepsen +3 more

mika03
# 
http://www.nabble.com/file/p16349649/at-length.png 

Hi,

I produced this graph with R and I am very happy with it. But now I wonder:
Is this some standard way of displaying data and if so, does it have a name?
(like Histogram, Box plot etc.)

I need to explain the plot to people so knowing a bit more about what
exactly this is, would certainly help ;-)

Thanks a lot!
Duncan Murdoch
#
On 3/28/2008 8:37 AM, mika03 wrote:
It's a stacked bar plot.  It's also somewhat similar to a mosaic plot, 
since the stacks all add up to 1, but in a mosaic plot, the rectangles 
typically have gaps between them, and don't necessarily have equal 
widths.  For example,

mosaicplot(~ Class + Survived, data = Titanic, color = TRUE)

Duncan Murdoch
Peter Jepsen
# 
Hi,

I think it looks like a spineplot.

Best regards,
Peter.

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On Behalf Of mika03
Sent: 28. marts 2008 13:37
To: r-help at r-project.org
Subject: [R] Quick question: Does this graph have a name?


http://www.nabble.com/file/p16349649/at-length.png 

Hi,

I produced this graph with R and I am very happy with it. But now I
wonder:
Is this some standard way of displaying data and if so, does it have a
name?
(like Histogram, Box plot etc.)

I need to explain the plot to people so knowing a bit more about what
exactly this is, would certainly help ;-)

Thanks a lot!
Hadley Wickham
# 
Which is a special case of a mosiac plot with only two variables.

Hadley

PS.  I suspect that the plot could be improved by ordering the x and y
variables by their marginal counts.  (Although maybe the y axis
already has an ordering?)
Frank E Harrell Jr
#
mika03 wrote:
A far better approach to this stacked bar chart would be a two-way dot 
chart which would also allow the inclusion of margins of error on the 
point estimates.
knussear
# 
I have always heard this referred to as a stack bar graph.

Ken
mika03 wrote:

  
    

      
      
    

  


  


      

    

    
2 days later

    

      
      

        


  

        

      mika03
    

    

      
      #
    

  


  

      
Thanks a lot for all your replies so far!

I now share your feeling that this might not be the best way to show the
data.

Here's what the data is about: We have search engine queries (12,000 of
them), which are grouped into semantic categories. (Some ask for a "Person",
some are about celebrity "Gossip" etc., you can see this on the x-axis.)

Then we asked three subjects for each query, how they think the query should
best be answered, we were mainly interested in the preferred length of the
response (y-axis: With a "phrase", a "sentence", a text "paragraph" etc.)

We want to show in this graph,
1) that users indeed think that queries from different semantic categories
should be answered with responses of different lengths. (That a Yes/No
answer basically.)
2) how the different length categories are distributed in the individual
semantic categories.

I hope this was somewhat clear..?

Thanks!