Need to tab multiple variables

[Jocelyn Cherry]

I wonder if someone can help? I am a new Stata user and have been struggling to find the solution to my problem in the help menus as I don't know the language to search with yet!

I have a large dataset of individuals with multiple questions answered by them. I am looking at the answers to ten questions on different types of injury- I have worked out the total number of 'yes' answers for each injury, but I need to make a table with the number of individuals who had any of the ten injuries. I have a unique identifier, but I cannot work out how to get Stata to give me the 'any injury' number.....I have tried by , list , tab, summarize with if as a condition and still am getting no joy.....help please!

[Paul Mueller]

Hi Jocelyn,

you have dichotomous variables for each injury and you want the number of people who gave at least one positive answer?

[Jocelyn Cherry]

I have yes, no, missing, withdrawn for each injury coded as 1,2, dot and 99....and yes, who gave at least one positive answer.

[Rich Goldstein]

you don't tell us what the varnames are but here are two ways:

assume the varnames are var1-var10:

gen byte anyyes=0
forval i=1/10 {
replace anyyes=1 if var`i'==1
}

then you can just tab anyyes

another way would be to use inlist as in

gen byte anyyes=inlist(1,var1, var2, etc)

the latter technique might require two inlists because the list of 10 might be too long

[Nick Cox]

tabm from tab_chi (SSC) may be what you seek. In your case, be sure to note the missing option.

ssc inst tab_chi

clear
set seed 2803
set obs 100

forval j = 1/10 {
    gen y`j' = runiform() > (`j') / 11
}

tabm y1-y10


. tabm y1-y10

           |        values
  variable |         0          1 |     Total
-----------+----------------------+----------
        y1 |         6         94 |       100
        y2 |         9         91 |       100
        y3 |        23         77 |       100
        y4 |        32         68 |       100
        y5 |        48         52 |       100
        y6 |        52         48 |       100
        y7 |        64         36 |       100
        y8 |        76         24 |       100
        y9 |        84         16 |       100
       y10 |        88         12 |       100
-----------+----------------------+----------
     Total |       482        518 |     1,000

[Rich Goldstein]

note that Nick and I clearly interpreted your question differently - you will have to decide which of us is "right"

[Jocelyn Cherry]

Thank you - am trying out both now!

[Nick Cox]

Rich is probably right here: I was reacting instinctively to your title, although your content is quite different. Nevertheless, tabm should still be invaluable!

With the same wild guesses about variable names,

egen any = anymatch(y1-y10), value(1)
su any

is another way to do it. In effect, it's a canned way of implementing what Rich did from first princoples.

[Jocelyn Cherry]

This is wonderful! This [Rich's answer] is exactly what I needed....I have been looking for so long through the help menus, thank you so much for taking the time to explain it to me.

[Jocelyn Cherry]

Oh hang on.....this doesn't get around my initial problem does it? This is a good way to get Stata to count the total number of injuries, but what I want is the total number of individuals that had an injury....I want to know the count of the unique identifiers with anyyes=1, not the count of anyyes.....how do I do that?

[Jocelyn Cherry]

The problem is one individual can have suffered 3 different injuries.

[Nick Cox]

Quite correct. Your #10 and #11 crossed with my edit as I realised my error.

Another answer is to look at help egen to see what is available. The two functions are similar and documented together.

[Rich Goldstein]

now I am confused; from you original post (#1), I had thought the data were in wide form and that is what I assumed when answering; however, I am now unclear about what your dataset looks like; if you would post (using CODE block delimiters - see the FAQ), a small amount of your data that would greatly help

note that if your data is in wide form (one row/person), then my original suggestion gives you want you want - the number of people with at least one injury