Fredrik Wallenberg
> I have a few graphs that I cannot seem to create using Stata 8.
> Perhaps someone could point me in the right direction (or confirm that
> it can't be done ... without creating a new dataset).
>
> I have two multiple choice survey questions. I would like to visualize
> the combinations of the answers in one out of three ways.
>
> A) Use a scatterplot but use the area of each coordinate dot to
> illustrate how many observations there are for that combination. The
> best I've done to date is using matrix and scatter, but that doesn't
> quite do it.
>
> graph matrix q3_3_4 q3_3_3, half jitter(3) msize(tiny)
> ... I can see how I could do A by creating a variable that had,
> as the weight, the total count for each combination and then us that
> as the aweight. Is there a nicer way?
As you say you could -- and should -- create a weight.
bysort q3_3_4 q_3_3 : gen N = _N
egen tag =tag(q_3_3_4 q_3_3)
scatter q_3_4 q_3_3 [w = N] if tag
> B) Use a 3 dimensional histogram (with the two variables and the x and
> z axis and the join distribution on the y axis). This can easily be
> done in Excel, why not in Stata?
Why not? Partly because most people who want to do this can do it easily
in Excel, I imagine... But it's clearly a reasonable thing to want.
> C) use -histogram, by-, BUT get ONE graph, treating the -by- options
> as separate variables in the same graph. Once again this seems like a
> reasonable thing to do but Stata doesn't have any support for it (that
> I can tell).
You mean you want them superimposed?
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/