Thought this was cool, since scatterplots with large n's are always a problem in SPSS. Just save it to read sometime.
Love you, Suz At 11:44 PM 11/14/01 -0500, you wrote: > >In my research methods course students have collected data so that they > >can compute one simple correlation (n=180). Both questions are on > >5-point scales. I want them to plot the data to see if it looks like > >there is a correlation. However, a simple scatter plot is uninformative > >because there is at least one subject who is reporting every combination > >on the 2 questions (so the scatterplot looks like a grid). Is there some > >way of plotting the data so that the "dots" are bigger when there are > >many cases. Or maybe one needs an entirely different plotting strategy. > >I use SPSS. Help? >========================= >One creative solution I ran across is this: create a "jittered" scatterplot. >Jittering simply moves dots to one side or the other, letting you see how many >there are. It's sort of like stacking pennies in several piles on a >table, then >shaking or jittering the table and having the pennies scatter near their >original spots. > Here's how to do it in SPSS: >Suppose you want to plot height vs. sex for N=100 males and females. Create a >new variable using COMPUTE: > sexj=rv.norm(sex,.1) [rv.norm is a function that can be picked off the >function list--it randomly and normally distributes the points] > now just do a scatterplot of height vs. sexj >You'll be able to see the number of points near each combination. >If you have two categorical variables, you could do the jittering in both >directions if you wish. > >Allen Shoemaker >Calvin College > > >--- >You are currently subscribed to tips as: [EMAIL PROTECTED] >To unsubscribe send a blank email to [EMAIL PROTECTED] --- You are currently subscribed to tips as: [EMAIL PROTECTED] To unsubscribe send a blank email to [EMAIL PROTECTED]