Merging multiple binary variables and hereby ignoring missing values

[Elise Sobrie]

Dear all

I have seven binary variables that I want to merge into one binary variable.
My seven variables (Rejected1, Rejected2, ..., Rejected7) all have value 1 if the firm is rejected for a credit, 0 otherwise.
Now I want to make RejectedMerge, the problem is that all the missing values of the seven variables are always included as 1 in RejectedMerge...
How do I make sure that they are excluded from the values 0 and 1 and thus that they just remain missing values?

I have already tried:
RejectedMerge = Rejected1 & Rejected2 &... & Rejected7 (also tried it with *)
egen RejectedMerge = group (Rejected1 Rejected2 ... Rejected7) (aslo used rowtotal)
but all of them include missing values

Thank you very much in advance!

[Abdul Adam]

Elise: your question is confusing. If you used dataex (as you are told in FAQ #12), things could have probably been easy to understand.
Having said that,
Is this what you are aiming for?

clear
set obs 3
input byte(rej1 rej2 rej3 rej4)
1 0 1 .
0 1 0 0
0 0 1 1
egen rej_tot = rowtotal(rej1 rej2 rej3 rej4)

Note: In the example, there are only 4 variables; but this can be extended to 7. rej_tot is not binary; but shows a total number of times a firm is rejected (2 for 1 & 3).

[Carlo Lazzaro]

Elise:
another approach might be:

. set obs 10
number of observations (_N) was 0, now 10

. g A=1 in 1/9
(1 missing value generated)

. g B=1 in 1/7
(3 missing values generated)

. foreach var of varlist A-B {
  2.
.                 replace `var'=999 if `var'==.
  3.
.         }
(1 real change made)
(3 real changes made)

. egen count=rowtotal(A B)

. replace count=count-999 if count>999
(3 real changes made)

. foreach var of varlist A-B {
  2.                 sum `var' if `var'!=999
  3.         }

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           A |          9           1           0          1          1

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           B |          7           1           0          1          1

However, please note that, -egen- with -rowtotal- function returns missing values as zeros:

. set obs 10
number of observations (_N) was 0, now 10

.
. g A=1 in 1/9
(1 missing value generated)

.
. g B=1 in 1/7
(3 missing values generated)

.
. egen count=rowtotal(A B)

. tab count

      count |      Freq.     Percent        Cum.
------------+-----------------------------------
          0 |          1       10.00       10.00
          1 |          2       20.00       30.00
          2 |          7       70.00      100.00
------------+-----------------------------------
      Total |         10      100.00

[Elise Sobrie]

Hallo

I guess I explained it somewhat ambiguous (I just find it hard to explain) because I don't want to count the frequencies..
I added a picture of a sketch of what I want to obtain. It is a kind of a list under the variable name Rm (RejectedMerged), so that each firm has now 7 observations related to the 7 types of credits. I don't know that what I want is actually possible in Stata...
If it is still not clear, please ask me

[Elise Sobrie]

Both, thank you very much for your effort!

Carlo, when I try your code I get an error:
. foreach var of varlist Rejected1 Rejected2 Rejected3 Rejected4 Rejected5 Rejected6 Rejected7 {
2. replace 'var' = 999 if 'var' ==.
3. }
' invalid name
r(198);

What is going wrong?

[Carlo Lazzaro]

Elise:
try again replacing

'var'

with

`var'

[Vanessa Owusu Ansah]

I want to create a categorical variable with binary variables which takes 1 if true and 0 otherwise so I used the syntax below.

egen byte new = rowtotal(childfever2 childcough5 underweight)

If I get 0,1,2 and 3 , what do they mean?

[Daniel Schaefer]

The number you get after rowtotal() is the number of 1's for a particular respondent on that particular observation. What that means substantively depends on the binary variables and what they mean. For example, each question could be part of a scale or index, and a greater value on the new variable might indicate a higher level of the underlying latent construct you want to measure. It depends on what your variables represent.