Chi-square goodness of fit ?

[Tony K]

Greetings,

I am new to STATA and I need help with adding categories to a variable. Here is what I have done so far.

I used ICD9 codes and extracted the observations that have the diagnosis that I need. For each of them I generated a new variable (CODE1, CODE 2 etc) that have the number of observations for that specific condition. Now I would like to compare these categories to see if there is a statistical difference by ethnicity across each diagnosis. I am trying to merge these CODE variables into a single variable that has all CODE listed as categories. I dont know how to proceed since using gen DXALL max(CODE 1, CODE 2 etc) just adds them together under 1 category.

My command for the test is this
csgof race, expperc (25 15 15 15 15 15)

Please let me know if there are any other details I should provide

[Nick Cox]

I don't understand all of what what you did, but it sounds like making a simple problem very hard. A chi-square test comes most often and most easily out of a cross-tabulation. See for example

 
. sysuse auto
(1978 Automobile Data)

. tab fore rep78, chi

You really shouldn't need to calculate frequencies yourself. csgof, which you should explain as coming from UCLA, is not for testing association, but for fitting a single distribution.

Please study FAQ Advice at http://www.statalist.org/forums/help, especially Sections 6, 12 and 18.

[Tony K]

Sorry I wasn't clear on my initial post and thank you for pointing me to the FAQ. I had to recode the variables because they were coded as strings. This is what generated label values for ICD9 codes that I need. I was using

 
 tab CODE1 RACE, chi2

But this doesnt provide me with percentages so I thought I'd use csgof just to get %

[Sarah Edgington]

Look at the help for tabulate twoway to find out more information about displaying percentages.
You'll want the row, column, or cell option, depending on which percentages you're interested in.

[Tony K]

Thank you for your help Sarah. Can you please tell me which help section should I look at if I want to merge categories to make a single variable with all categories listed?

[Nick Cox]

Being string is no barrier to the process I cited. The only difference is whether missings are included by default. Try

 
. sysuse auto
(1978 Automobile Data)

. tostring rep78 foreign, gen(s_rep78 s_foreign)
s_rep78 generated as str1
s_foreign generated as str1

. tab s_*, chi

While percents may be of interest for description, they aren't part of chi-square testing, which here is based on frequencies.