Categorical data - Graphical Representation

[Hajirah Kausar]

Hello,

I have three categorical variables and I wanted some sort of graphical representation for my thesis.
My topic is association between asthma and anxiety over the time period of 2012-2014.
My variables look like this.

. tab ConDr

Doctor diagnosed asthma | Freq. Percent Cum.
------------------------+-----------------------------------
Yes | 1,199 19.90 19.90
No | 4,825 80.10 100.00
------------------------+-----------------------------------
Total | 6,024 100.00

. tab GHQg2

(D) GHQ Score - grouped |
(0,1-3,4+) | Freq. Percent Cum.
------------------------+-----------------------------------
Score 0 | 3,229 58.29 58.29
Score 1-3 | 1,431 25.83 84.12
Score 4+ | 880 15.88 100.00
------------------------+-----------------------------------
Total | 5,540 100.00

. tab SYear

Survey year | Freq. Percent Cum.
------------+-----------------------------------
2012 | 1,964 32.58 32.58
2013 | 2,143 35.54 68.12
2014 | 1,922 31.88 100.00
------------+-----------------------------------
Total | 6,029 100.00

I wonder if someone could help me out with the appropriate stata command.
I tried this command gr bar GHQg2, by(ConDr SYear).

However I'd like the three categories of GHQg2 also taken into account. Not Its mean value.

Thank you.

[Nick Cox]

Help us by showing the results of the 3 x 3 x 2 breakdown. Showing the results of this would suffice:

ssc inst groups
groups ConDr GHQg2 SYear

[Hajirah Kausar]

Hello,

I tried your command
Got the following response.

. ssc inst groups
connection timed out -- see help r(2) for troubleshooting
http://fmwww.bc.edu/repec/bocode/g/ either
1) is not a valid URL, or
2) could not be contacted, or
3) is not a Stata download site (has no stata.toc file).

I'm using STATA 12.
Can you guide some other way.

Thank you.

[Nick Cox]

groups is there on SSC and accessible. I just checked. So, the problem looks like your internet connection. See also help netio

Alternatively, show us the results of

contract ConDr GHQg2 SYear
list

[Hajirah Kausar]

Thank you.

[Nick Cox]

Sorry, but that's an image and I can't copy and paste usefully.

http://www.statalist.org/forums/help#stata explains,

Use dataex (SSC) or -- if you can't download that -- copy and paste the list output to between CODE delimiters.

[Hajirah Kausar]

. contract ConDr GHQg2 SYear

. list

+----------------------------------+
| SYear ConDr GHQg2 _freq |
|----------------------------------|
1. | 2012 Yes Score 0 195 |
2. | 2013 Yes Score 0 239 |
3. | 2014 Yes Score 0 184 |
4. | 2012 Yes Score 1- 84 |
5. | 2013 Yes Score 1- 111 |
|----------------------------------|
6. | 2014 Yes Score 1- 97 |
7. | 2012 Yes Score 4+ 72 |
8. | 2013 Yes Score 4+ 74 |
9. | 2014 Yes Score 4+ 70 |
10. | 2012 Yes . 22 |
|----------------------------------|
11. | 2013 Yes . 26 |
12. | 2014 Yes . 25 |
13. | 2012 No Score 0 872 |
14. | 2013 No Score 0 906 |
15. | 2014 No Score 0 830 |
|----------------------------------|
16. | 2012 No Score 1- 341 |
17. | 2013 No Score 1- 421 |
18. | 2014 No Score 1- 377 |
19. | 2012 No Score 4+ 221 |
20. | 2013 No Score 4+ 229 |
|----------------------------------|
21. | 2014 No Score 4+ 214 |
22. | 2012 No . 155 |
23. | 2013 No . 136 |
24. | 2014 No . 123 |
25. | 2012 . Score 0 2 |
|----------------------------------|
26. | 2014 . Score 0 1 |
27. | 2013 . . 1 |
28. | 2014 . . 1 |
+----------------------------------+

.
end of do-file

[Nick Cox]

I can use that, thanks, but CODE delimiters are not difficult! Please do see the link given.

I'm assuming that diagnosis is the response. With tabplot installed, I engineer your listing and get this:

clear
set scheme s1color
input SYear str3 ConDr str2 GHQg2 _freq
2012 Yes 0 195
2013 Yes 0 239
2014 Yes 0 184
2012 Yes 1- 84
2013 Yes 1- 111
2014 Yes 1- 97
2012 Yes 4+ 72
2013 Yes 4+ 74
2014 Yes 4+ 70
2012 Yes . 22
2013 Yes . 26
2014 Yes . 25
2012 No 0 872
2013 No 0 906
2014 No 0 830
2012 No 1- 341
2013 No 1- 421
2014 No 1- 377
2012 No 4+ 221
2013 No 4+ 229
2014 No 4+ 214
2012 No . 155
2013 No . 136
2014 No . 123
2012 . 0 2
2014 . 0 1
2013 . . 1
2014 . . 1
end

replace ConDr = "" if ConDr == "."
replace GHQg2 = "" if GHQg2 == "."

tabplot ConDr SYear [fw=_freq], by(GHQg2, row(1) compact) percent(SYear GHQg2) showval

See http://www.statalist.org/forums/foru...updated-on-ssc
for more on tabplot.

With your original data you should be able to do this directly with

set scheme s1color
tabplot ConDr SYear, by(GHQg2, row(1) compact) percent(SYear GHQg2) showval

On the other hand, something like

graph bar (mean) ConDr, over(SYear) by(GHQ, row(1))

should show you fraction diagnosed in a direct way.

[Hajirah Kausar]

Thank you for being patient with me but I couldn't still get results.

. replace ConDr = "" if ConDr == "."
type mismatch
. tabplot ConDr SYear [fw=_freq], by(GHQg2, row(1) compact) percent(SYear GHQg2) showval
unrecognized command: tabplot
.
. tabplot ConDr SYear, by(GHQg2, row(1) compact) percent(SYear GHQg2) showval
unrecognized command: tabplot

. graph bar (mean) ConDr, over(SYear) by(GHQ, row(1)
parentheses do not balance

Much Regards.

[Hajirah Kausar]

The graph bar works. Thank you very much.

[Hajirah Kausar]

But your graph is better but I can't reproduce it.

[Rich Goldstein]

note that tabplot is a user-written command and must be downloaded and installed - use "search tabplot" to find and get instructions for a standard download

[Nick Cox]

Working backwards:

Rich in #12: Thanks for underlining that tabplot is user-written.

Hajirah: The link in #9 does make that explicit. When we provide links, you are expected to read them!

Hajirah in #10 You left off a parenthesis. Look again at #9.

graph bar (mean) ConDr, over(SYear) by(GHQ, row(1))

Some of the code in my answer is just based on my engineering your results because you did not provide a data example in the requested form. Note again the important proviso "With your original data you should be able to do this directly with <code follows>".
But looking at ConDr results again, I sense a poor coding. With a diagnosis variable codes of 1 for Yes and 0 for No are most convenient. Then the mean over 0s and 1s is just the fraction diagnosed, which makes clinical and scientific and statistical sense, or so I suggest.

[Hajirah Kausar]

Thank you. I'll write my queries properly next time. Sorry for the inconvenience.
Much Regards.

[Hajirah Kausar]

Hello again.

I wanted to ask if other than the variables ConDr, SYear and GHQg2 variables defined above , I could make a graphical representation of my data including another categorical variable as well (T).


. tab T

T Freq. Percent Cum.
------------------------+-----------------------------------
Yes | 959 53.10 53.10
No | 847 46.90 100.00
------------------------+-----------------------------------
Total | 1,806 100.00


Thank you.