[EpiData-list] multiple response (& server is up again)

epidata-list at lists.umanitoba.ca epidata-list at lists.umanitoba.ca
Tue Oct 7 15:48:19 CDT 2008


The server is now up and running again.
(All scripts have been checked - there were no problems, but only 
assumed problems).

Off list the following question was asked regarding EpiData Analysis
 > Hi There,
 >  I have a question that i am trying to evaluate with a multi response 
answer.
 >   I have set out the question as follow with a text variable for each 
answer:
 > 
 > Q1:  What fruit do you like _ _ _ _ _
 >  a - Apples  b - Orange  c - Banana  d - Pear  e - Grapes
 >  Now when i go to analyse this, is there a way i can graph the response
 > for this answer on a bar graph?  I cannot seem to be able to add
 > multiple series into the graph to compare all answers for this question?

Currently bar graphs can only take one variable at a time. Except for 
box and scatter/line plots for several variables. Box plots for several 
variables has just been introduced with the pre v2.1 - v2.0.5.148 version.
For all graphs a /by variable has been introduced, but this does not 
change the basic limitation of one variable per plot.

What you could do to analyse the data is to use stattables and count 
binary indicators calculated like this:
gen i apple = (pos(q1, "a" ) > 0)
gen i orange =  (pos(q1, "b" ) > 0) etc for the other fruits.

stattable apple orange ......../stat="sum" /N /close
* then in the new file with counts of the fruits and the total some 
percentages could be worked out.

A graph showing the combinations of the variables in the form of a VENN 
diagram could be a solution, but currently this cannot be done with 
EpiData Analysis. I wrote a command for Stata for this:
See: http://www.stata.com/products/stb/journals/stb47.pdf . The initial 
version has been extended with four factor ellipses.
We could consider including Venn diagrams in a later version (but at the 
earliest second quarter 2009) and also a multiple response tables command.

regards
Jens Lauritsen
EpiData Association


More information about the EpiData-list mailing list