Crosstabs of 2 categorical variables by group

[Andreas Head]

Hi everyone,
I am currently working on a large dataset with individual level data from three countries (in one dataset) using Stata 12.
Let's say I have a country identifier variable, one var that's educational attainment (4 categories plus one don't know" cat), and one that's attitude towards homosexuals (three categories 1=No; 2=yes, 3=don't know).

I need to create a table as follows:
for each country, tabulate only those frequencies of attitudes towards homosexuals of category 2(=yes), by educational attainment.

Based on such a table, I want to be able to easily read, for example, how many percentages of respondents with post-secondary qualifications state 2=yes in the attitudes variable.

Is there an easy way to do that?

Thanks,
Andreas

[Charlie Joyez]

bysort country : tab educational homosexual, col nofreq

Will give you for each country the percentage of each categories of attitude toward homosexual by educational level.
You can even recode the attitude variable into a binary (yes/no) to only have two alternatives by education level.


It would be easier to check whether this is what you want if you post some of your data using dataex (SSC).

This isn't yet in one table, you could manage to build you own table either by hand (fastidious), either using tabout (SSC). See help tabout for details

Best,
Charlie

[Andreas Head]

Thanks Charly,
I am aware of your suggested syntax. That's the way I ussually do. I just thought there might be a more "fancy" solution to my issue, by telling Stata to display exactly and only the information that I need for my table (and finally the respective graph I want to plot).
I tried

tabout

on this matter, but I couldn't get the appropriate syntax. Any suggestion on that matter?

Thanks

[Andreas Head]

Here is a similar graph to the one I want to create based on the stata output table:

You will notice that the variables are on the left hand side, the percentages display the percentage of respondents who answered with "yes", and the legend displays the three countries.

[Charlie Joyez]

I don't know exactly how to end up with the precise table you want, directly with one Stata line of code, this is why I let you browse the tabout helpfile.
I was thinking about running the previous command for each country, and use the append otpion in tabout to merge them in one table.
Maybe other people around here could help you doing this.

However, why do you want such a complicated table if your real objective is to draw graphs?
I would know a (long and certainly not the most efficient) way to simply collect each of these proportions into adequate new variables, and then draw the graphs, skipping the table intermediate.

But perhaps you really want that table...

Best,
CharlIE

[Andreas Head]

Thanks, Charly!

Well I will have to make the graph using excel. That is why I need a table in stata in the first place.

However, why do you want such a complicated table if your real objective is to draw graphs?
I would know a (long and certainly not the most efficient) way to simply collect each of these proportions into adequate new variables, and then draw the graphs, skipping the table intermediate.

Would you kindly share the way you do that?

Thanks

[Charlie Joyez]

Actually this was easier than I first thought,
but this remains Stata graphs.

Edit : for a generalization to a larger number of countries, I used the distinct command (by Cox & Longton, from SSC).

clear
input byte attitude float country byte educational
1 1 5
2 1 1
1 1 3
2 1 2
3 1 1
2 1 2
1 1 2
2 1 2
3 1 2
1 1 3
3 1 1
3 1 1
3 1 1
1 1 2
2 1 1
2 1 2
2 1 2
1 1 2
1 1 1
1 1 4
1 1 1
2 1 1
3 1 3
1 1 4
1 1 3
3 1 1
3 1 1
3 1 1
1 1 2
2 1 2
3 1 1
2 1 1
3 1 2
1 1 3
3 1 1
2 1 2
2 1 2
2 1 2
1 1 3
1 1 2
3 1 1
2 1 3
1 1 3
1 1 4
3 1 4
1 1 4
2 1 2
2 1 3
2 1 2
2 1 2
1 1 3
1 1 2
3 2 4
3 2 5
3 2 5
3 2 5
2 2 5
2 2 4
3 2 5
1 2 4
2 2 3
2 2 4
1 2 5
3 2 5
1 2 5
1 2 5
3 2 3
1 2 4
2 2 3
3 2 5
2 2 5
2 2 5
3 2 5
3 2 3
end


 bysort country edu: gen yes=(att==2)

 collapse yes, by(country educ)

 
distinct country
local ncountry=r(ndistinct)
forvalues c= 1/`ncountry'{
graph hbar yes if country==`c',over(educ)
graph save graph`c' ,replace
}
graph combine "graph1" "graph2"

[Nick Cox]

Charlie: Thanks for the mention of distinct, but the Stata Journal website carries the latest version. Either way,

search distinct

in Stata will call up references, notably to the longer discussion by Gary Longton and myself.

[Charlie Joyez]

Also, see

help graph bar

To see how to add labels or colors in your graphs and

help graph combine

To see how to change the display of the combined graphs.

[Charlie Joyez]

Nick Cox
Ok, I take good note of that.
Actually I didn't had the latest version of your program then, which is now corrected.
I'll take immediately a look at the changes. (glad to see the r(ndistinct) scalar is still here, so my previous code would work).

Best,
Charlie

[Andreas Head]

Thank you, Charly and Nick.

I will try your code with -distinct-.

[Nick Cox]

Indeed, as I have mentioned in another thread, I don't break previous versions without really good reasons!

Here's another approach to the graphics. I use designplot, first posted on SSC but now to be obtained from the Stata Journal website.

The original post announcing the program was at

http://www.statalist.org/forums/foru...riptive-tables

and there have been further mentions since in this forum.

In Stata

search designplot

will bring up a reference to a 2014 paper (accessible with a subscription or a one-off payment) and to the 2015 software update. The software should be installed by first clicking on gr0061_1.

SJ-15-2 gr0061_1 . . . . . . . . . . . . . . . Software update for designplot
(help designplot if installed) . . . . . . . . . . . . . . N. J. Cox
Q2/15 SJ 15(2):605--606
bug fixed for Stata 14

SJ-14-4 gr0061 Design plots for graphical summary of a response given factors
(help designplot if installed) . . . . . . . . . . . . . . N. J. Cox
Q4/14 SJ 14(4):975--990
produces a graphical summary of a numeric response variable
given one or more factors

(Incidentally, the Stata Journal team know of the small gremlin whereby an incorect title "The chi-square goodness-of-fit test for count data models" is displayed. That will be fixed as soon as possible.)

designplot yes country educational, variablenames t1title(Fraction saying Yes) ysc(r(-0.02 .)) marker(1, ms(O)) yla(0 "0" 0.1(0.1)0.6, format(%02.1f)) ytitle("")

I used Charlie's useful fake data posted in #7. The example here is at best indicative, not definitive. I have a preference here for dot charts over bar charts, especially given that two means are exact zeros, which has to be explained otherwise.

The main idea is that you can show results at various levels of aggregation, here for all the data, for two countries, for five educational levels, and for the cross-combinations that occur.