For others who might be interested, I got an email from this person.
Usually, I refer people to the newsgroups or just delete the email.
This time I took the time to actually respond. Below is a copy of that
reply.
I assume you mean surveys that were conducted pre and post some event
(such as a class) rather than pre-survey and post-survey because the
latter, at least to me, is meaningless.
There are a variety of ways to plot this, all being on the same chart.
If your data are normally distributed plot both sets. I believe you
should also be able to test the probability of getting the post-event
result even if the pre-event conditions hold.
If you have data subject by subject, plot them as such (subject on the
x-axis, the results in 2 series on the y-axis).
You could also plot the pre- data as the x-values and the post-data on
the y-axis in a XY Scatter chart. I don't know the statistical
validity of then using the slope to reach conclusions about the value
of the event.
Search google for more ideas, including
http://teachhealthk-
12.uthscsa.edu/pa/pa11/11.htm
As visually appealing all these may be, the real test would be a
statistical test, probably something like a chi-square to test if the 2
samples are truly statistically different.
--
Regards,
Tushar Mehta
www.tushar-mehta.com
Excel, PowerPoint, and VBA add-ins, tutorials
Custom MS Office productivity solutions
