recode string as numeric?

[River Huang]

Suppose I have the following data:

* Example generated by -dataex-. To install: ssc install dataex
clear
input str5 t str1 gz
"C    " "Y"
"A++  " "N"
"B    " "N"
"A+   " "N"
"C    " "Y"
"A    " "N"
"A-   " "N"
"C    " "N"
"A    " "N"
"-    " "Y"
"C    " "N"
"B    " "N"
"C-   " "N"
end

1. For variable t, I have 7 categories, e.g., A++ (1), A+ (2), A (3), A- (4), B (5), C (6), and C- (7). Missing data are denoted as "-". I want to recode the string variable t into numerical variable, say t1, with values denoted in the corresponding parenthesis, and "-" are replaced by the standard ".".
2. Similarly, for gz variable, I want to construct a numeric variable gz1 with "N"/"Y" replaced by 0/1, respectively.
Any suggestions?

[Friedrich Huebler]

gen gz1 = 0
replace gz1 = 1 if gz=="Y"

Same for t, only with more replace commands.

[Charlie Joyez]

Or,

. encode t,gen(t_code)

. replace t_code=. if t=="-"

[Friedrich Huebler]

. encode t,gen(t_code)
. replace t_code=. if t=="-"

These commands don't yield the values that River needs.

. lab list t_code
t_code:
           1 -    
           2 A    
           3 A+   
           4 A++  
           5 A-   
           6 B    
           7 C    
           8 C-

[Marcos Almeida]

I suggest the following commands:

. encode t, gen(t1)

. replace t1 = . if t1 ==1
(1 real change made, 1 to missing)

. codebook t1

----------------------------------------------------------------------------------------------------------------------------------------------
t1                                                                                                                                 (unlabeled)
----------------------------------------------------------------------------------------------------------------------------------------------

                  type:  numeric (long)
                 label:  t1

                 range:  [2,8]                        units:  1
         unique values:  7                        missing .:  1/13

            tabulation:  Freq.   Numeric  Label
                             2         2  A
                             1         3  A+
                             1         4  A++
                             1         5  A-
                             2         6  B
                             4         7  C
                             1         8  C-
                             1         .  

. list

     +------------------+
     |     t   gz    t1 |
     |------------------|
  1. | C        Y     C |
  2. | A++      N   A++ |
  3. | B        N     B |
  4. | A+       N    A+ |
  5. | C        Y     C |
     |------------------|
  6. | A        N     A |
  7. | A-       N    A- |
  8. | C        N     C |
  9. | A        N     A |
 10. | -        Y     . |
     |------------------|
 11. | C        N     C |
 12. | B        N     B |
 13. | C-       N    C- |
     +------------------+

With regards to the second query, the command may be done within one line as well:

. gen gz1 = gz =="Y"

. list

     +------------------------+
     |     t   gz    t1   gz1 |
     |------------------------|
  1. | C        Y     C     1 |
  2. | A++      N   A++     0 |
  3. | B        N     B     0 |
  4. | A+       N    A+     0 |
  5. | C        Y     C     1 |
     |------------------------|
  6. | A        N     A     0 |
  7. | A-       N    A-     0 |
  8. | C        N     C     0 |
  9. | A        N     A     0 |
 10. | -        Y     .     1 |
     |------------------------|
 11. | C        N     C     0 |
 12. | B        N     B     0 |
 13. | C-       N    C-     0 |
     +------------------------+

[Daniel Bela]

In addition to Marcos Almeida's excellent advice, you should have a look at -encode-'s option -label()-: If you want to have control over the numeric codes used by encode, you can (1) define a value label containing the desired codes and (2) let encode use this value label.

* Example generated by -dataex-. To install: ssc install dataex
clear
input str5 t str1 gz
"C    " "Y"
"A++  " "N"
"B    " "N"
"A+   " "N"
"C    " "Y"
"A    " "N"
"A-   " "N"
"C    " "N"
"A    " "N"
"-    " "Y"
"C    " "N"
"B    " "N"
"C-   " "N"
end

label define ratings 1 "A++" 2 "A+" 3 "A" 4 "A-" 5 "B" 6 "C" 7 "C-" .a "-"
replace t=trim(itrim(t))
encode t , generate(t_code) label(ratings) noextend

label define yesno 1 "Y" 2 "N"
replace gz=trim(itrim(gz))
encode gz , generate(gz_code) label(yesno) noextend

Note that, for the sake of keeping everything a little easier, my code starts with trimming blank characters off the string values of your original variables. The option -noextend- makes -encode- stop execution in case any values not defined in the given value label are encountered in the source string variable.

Regards
Bela

[River Huang]

In addition to Marcos Almeida's excellent advice, you should have a look at -encode-'s option -label()-: If you want to have control over the numeric codes used by encode, you can (1) define a value label containing the desired codes and (2) let encode use this value label.

Thank you all for the suggestions. @Friedrich: I knew your suggestive method, but tried to search for a more concise approach. @Charlie, @Friedrich and ＠Marcos：I also knew the "encode" command can do something like this. But the results are NOT what I had in mind (say, I wanted A++ to be 1 rather than 4). @Daniel's answer is exactly what I need. Thank you, Daniel. Thank you all.

[Marcos Almeida]

Thank you all for the suggestions. @Friedrich: I knew your suggestive method, but tried to search for a more concise approach. @Charlie, @Friedrich and ＠Marcos：I also knew the "encode" command can do something like this. But the results are NOT what I had in mind (say, I wanted A++ to be 1 rather than 4)

I presume you also knew how "to construct a numeric variable gz1 with "N"/"Y" replaced by 0/1, respectively."

No doubt, Daniel Bela shared excellent commands. What is more, he envisaged issues which could well bite, and gave excellent strategy on how to use the -noextend- option. This was a good lesson I learned today. Thanks, Daniel.

That said, as you know, what matters most when categorizing a variable is providing the correct label. However much we further change the codes. the results will keep the same, provided the label is correct.

This notwithstanding, for the exact "end" result you wished, I mean, the command, as you may also know, - recode - will simply do fine:

 . gen t2 = t1
(1 missing value generated)

. recode t2 (2=1 "A") (3=2 "A+") (4=3 "A++") (5=4 "A-") (6=5 "B") (7=6 "C") (8=7 "C-") (miss=.), gen(t3)
(12 differences between t2 and t3)

. codebook t3

----------------------------------------------------------------------------------------------------------------------------------------------
t3                                                                                                                                RECODE of t2
----------------------------------------------------------------------------------------------------------------------------------------------

                  type:  numeric (float)
                 label:  t3

                 range:  [1,7]                        units:  1
         unique values:  7                        missing .:  1/13

            tabulation:  Freq.   Numeric  Label
                             2         1  A
                             1         2  A+
                             1         3  A++
                             1         4  A-
                             2         5  B
                             4         6  C
                             1         7  C-
                             1         .

As previously said, the result - whatsoever it is - after the second labeling - won't change:

. tab t1

         t1 |      Freq.     Percent        Cum.
------------+-----------------------------------
      A     |          2       16.67       16.67
      A+    |          1        8.33       25.00
      A++   |          1        8.33       33.33
      A-    |          1        8.33       41.67
      B     |          2       16.67       58.33
      C     |          4       33.33       91.67
      C-    |          1        8.33      100.00
------------+-----------------------------------
      Total |         12      100.00

. tab t3

  RECODE of |
         t2 |      Freq.     Percent        Cum.
------------+-----------------------------------
          A |          2       16.67       16.67
         A+ |          1        8.33       25.00
        A++ |          1        8.33       33.33
         A- |          1        8.33       41.67
          B |          2       16.67       58.33
          C |          4       33.33       91.67
         C- |          1        8.33      100.00
------------+-----------------------------------
      Total |         12      100.00

. tab t1 t3

           |                                 RECODE of t2
        t1 |         A         A+        A++         A-          B          C         C- |     Total
-----------+-----------------------------------------------------------------------------+----------
     A     |         2          0          0          0          0          0          0 |         2
     A+    |         0          1          0          0          0          0          0 |         1
     A++   |         0          0          1          0          0          0          0 |         1
     A-    |         0          0          0          1          0          0          0 |         1
     B     |         0          0          0          0          2          0          0 |         2
     C     |         0          0          0          0          0          4          0 |         4
     C-    |         0          0          0          0          0          0          1 |         1
-----------+-----------------------------------------------------------------------------+----------
     Total |         2          1          1          1          2          4          1 |        12

[River Huang]

I presume you also knew how "to construct a numeric variable gz1 with "N"/"Y" replaced by 0/1, respectively."

No doubt, Daniel Bela shared excellent commands. What is more, he envisaged issues which could well bite, and gave excellent strategy on how to use the -noextend- option. This was a good lesson I learned today. Thanks, Daniel.

That said, as you know, what matters most when categorizing a variable is providing the correct label. However much we further change the codes. the results will keep the same, provided the label is correct.

This notwithstanding, for the exact "end" result you wished, I mean, the command, as you may also know, - recode - will simply do fine:

Many thanks again, Marcos. I suppose that you meant

recode t2 (2=3 "A") (3=2 "A+") (4=1 "A++") (5=4 "A-") (6=5 "B") (7=6 "C") (8=7 "C-") (miss=.), gen(t3)

which is what I needed. Your suggestion of using "recode" in this way is even more intuitive. I like it. Thanks.