Graphical Correlation between Discrete X,Y
Six Sigma – iSixSigma › Forums › Old Forums › General › Graphical Correlation between Discrete X,Y
 This topic has 7 replies, 7 voices, and was last updated 19 years, 6 months ago by John Noguera.

AuthorPosts

May 13, 2003 at 3:53 pm #32228
Folks
I have a case of a discrete X and a discrete Y I am interested in representing graphically correlationships. For example Salesmen (5) against Product Source (5 Warehouses). I have already used pivot tables and stackup histograms…but want a better graphical representation. Any suggestions?0May 13, 2003 at 5:18 pm #85847
Ed VanHauteParticipant@EdVanHaute Include @EdVanHaute in your post and this person will
be notified via email.If there is a relationship between the salesmen and the warehouses, you can use a dot plot with a regression equation included. I am guessing, however, that the salesmen and the warehouses are both X’s, and you have yet to identify the Y. When you do, use a matrix plot, multivari analysis, and/or a multiple regression analysis toshow the relationship between your inputs and outputs
0May 14, 2003 at 5:21 am #85860Thanks for your response. I do know the Y and the Y is Delivery Time. I agree with you on both being Xs but this is only to establish a relationship. Also just for you to know the salesman as per a moods median test (nonnormal data) is not a critical X. I am still not clear of using a regression equation for discrete Xs.
0May 14, 2003 at 1:29 pm #85874You could run a correlation/covariance analysis to see whether or not your Xs are correlated. A scatterplot of each X versus another will show that if you need it graphically.
As far as running a regression analysis on discrete X I am going to assume that you really have multiple level categorical data which means you need to run a logistic regression. If your data has order to it, such as a Likert scale you should use Ordinal Logistic Regression. If Salesman A could swich places with Salesman B, ie call A B and call B A, you should use nominal logistic regression. What the correlation analysis will tell you is which two level factors need to be included when you input the model. What it won’t tell you is which higher order interactions (three and four way) need to be included.0May 29, 2003 at 5:42 am #86454
SwaggertyParticipant@George Include @George in your post and this person will
be notified via email.Here’s a simple graph: square the correlation to get the shared variance between X & Y. Multiply by 100 to get a percentage. Plot the percentages in a bar chart and call it “Leverage”. E.g., this is what salesman X needs to leverage.
0May 29, 2003 at 2:59 pm #86472Use a radar chart (aka spider chart, star plot). The warehouses will be on the outside edge of the chart, with each as one edge of a pentagon. Results for each of the five salesmen at each warehouse will be plotted on the inside of the chart with their results colorcoded by salesman and the dot of each salesman at each warehouse connected.
The result is a single, datarich graph that will show:how each salesman performed at each warehouse
how each salesman performed relative to each of the other salesmen at each warehouse
the overall ranking of the salesmen (which is the total area of the graph enclosed by their connected points).
Here is additional information:
http://www.itl.nist.gov/div898/handbook/eda/section3/starplot.htm
http://relish.concordia.ca/Quality/tools/23radar.pdf0May 29, 2003 at 11:00 pm #86491
JonathonParticipant@Jonathon Include @Jonathon in your post and this person will
be notified via email.You mention pivot tables, but have you tried a pivot chart? They are pretty powerful, especially if you create a dualaxis chart. That is, the chart holds an X1 and X2 as variables, and displays “towers” for each X1X2 comnbination. It’s pretty cool stuff, but you need Office 2000 or later.
0May 30, 2003 at 11:59 am #86504
John NogueraParticipant@JohnNoguera Include @JohnNoguera in your post and this person will
be notified via email.Hi Jonathon!
How’s the book coming along?
I wanted to mention that SigmaXL creates pivot charts for Excel ’97.
Also, I wanted to mention that SAS JMP has a similar powerful tool for Discrete X and Discret Y that they call Mosaic Plots.0 
AuthorPosts
The forum ‘General’ is closed to new topics and replies.