How to replace any variable's value if meeting certain condition

[Chen Samulsion]

Dear Stata users,

In the following data, I want to replace variables d2a1 d2a2 d2a3 to missing 'cause variable d2a <100. Now I must type as many command lines as certain conditions.

replace d2a1=. if d2a==29
replace d2a1=. if d2a==8
replace d2a2=. if d2a==50
replace d2a3=. if d3a==50
......

Or:

replace d2a1=. if d2a<100
replace d2a2=. if d2a<100
replace d2a3=. if d3a<100

How to achieve that in a smarter way? (one command line and not so many "if", for example) Thank you.

    d2a1      d2a2      d2a3          d2a4        d2a
      29         0         0          2500         29
       8         0         0             0          8
       0        50         0         20000         50
      50         0         1          5000         50
       0         1         0         30000          1
       0        21         0         30000         21
      50         0         7          5000         57
      10        50         0         20000         60
      29         0         1          2500         30

[daniel klein]

I do not see any obvious pattern in your first example.

The second example can be wrapped into a loop

foreach var in d2a1 d2a2 d2a3 {
    replace `var' = . if d2a < 100
}

or

forvalues i = 1/3 {
    replace d2a`i' = . if d2a < 100
}

[Chen Samulsion]

There are some typo in 1#, d3a for example. I have used loop before opening this thread but expect a smarter way. Maybe loop is smart enough. Thank you daniel klein.

[daniel klein]

I have used loop before opening this thread but expect a smarter way.

How do you define "smart" here? If you plan on doing these things more often, define a program

program replace_varlist
    version 16.1
    syntax varlist =exp [ if ] [ in ] [ , *]
    foreach var of local varlist {
        `var' `exp' `if' `in' , `options'
    }
end

that you then call as

replace_varlist d2a1 d2a2 ...

Obviously and depending on details, d2a1, d2a2, ... may be referred to as d2a*?.

[Chen Samulsion]

Thank you very much daniel. Well, in my data, d2a1+d2a2+d2a3 should be equal to 100 (d2a, the sumup variable). If d2a !=100, suggesting that observations are recorded wrongly and should be set to missing. Previously, I prepared to replace the trouble-maker variable (d2a1 in 18th observation, for example) to missing. So I want to find a "smart" way that help me finding the trouble-maker variable in each observation. Now, I will replace all three variables to missing if the sumup variable is not equal to 100, so the "smart" way seems not be badly in need of.

. foreach v in d2a d2b d2c d2d d2e {
  2.  list `v'1 `v'2 `v'3 `v' if `v'<100 & `v'!=0
  3.  }

      +--------------------------+
      | d2a1   d2a2   d2a3   d2a |
      |--------------------------|
  18. |   29      0      0    29 |
 564. |    8      0      0     8 |
1150. |    0     50      0    50 |
1256. |   50      0      0    50 |
1426. |    0      1      0     1 |
      +--------------------------+

      +--------------------------+
      | d2b1   d2b2   d2b3   d2b |
      |--------------------------|
  18. |   71      0      0    71 |
1150. |    0     50      0    50 |
1256. |   50      0      0    50 |
      +--------------------------+

      +--------------------------+
      | d2c1   d2c2   d2c3   d2c |
      |--------------------------|
 214. |   33     33     33    99 |
1214. |    0      1      0     1 |
1697. |    0      5      0     5 |
      +--------------------------+

      +--------------------------+
      | d2d1   d2d2   d2d3   d2d |
      |--------------------------|
1214. |    0      1      0     1 |
      +--------------------------+

      +--------------------------+
      | d2e1   d2e2   d2e3   d2e |
      |--------------------------|
 736. |    1      0      0     1 |
1214. |    0      3      0     3 |
1310. |    8      0      0     8 |
1465. |    1      1      1     3 |
1697. |    1      0      0     1 |
      |--------------------------|
1874. |    0      0      8     8 |
      +--------------------------+

[daniel klein]

Well, in my data, d2a1+d2a2+d2a3 should be equal to 100 (d2a, the sumup variable). If d2a !=100, suggesting that observations are recorded wrongly and should be set to missing.

It could also suggest that correctly recorded variables have been added up incorrectly.

[...] So I want to find a "smart" way that help me finding the trouble-maker variable in each observation.

How do you know (i) how many "trouble-maker" variables there are in a given observation, and (ii) which of potentially many variables is the "trouble-maker"? I think you ask for the impossible. However, I obviously have no background information whatsoever, to be sure.

[Chen Samulsion]

The added up variable d2a is generated manually by code: egen d2a=rowtotal(d2a1 d2a2 d2a3), so it is no possible of incorrectness.
To know how many "trouble-maker" variables there are in a given observation is impossible, so I will replace variables to missing whenever their corresponding added up variable is not equals to 100.
This thread can be closed by now.