Changing variable's value by frequency order

[Jean Jacques]

Hi, i have a variable x with 5 values (assume 1 to 5). When I do tab x I have that the value with higher frequency is 2 and the next value with higher frequency is 5. I want to recode 2 as 1, 5 as 2, and so on for all the values of my variable. Clearly I have to do that with many variables, that's why I cannot do it manually.

[Nick Cox]

Why not just use the sort option?

. webuse nlswork, clear
(National Longitudinal Survey.  Young Women 14-26 years of age in 1968)

.. tab occ_code

 occupation |      Freq.     Percent        Cum.
------------+-----------------------------------
          1 |      3,008       10.59       10.59
          2 |      1,494        5.26       15.84
          3 |     10,974       38.62       54.47
          4 |      1,323        4.66       59.12
          5 |        438        1.54       60.67
          6 |      4,309       15.17       75.83
          7 |        571        2.01       77.84
          8 |      4,300       15.13       92.98
          9 |          6        0.02       93.00
         10 |        144        0.51       93.50
         11 |        194        0.68       94.19
         12 |          7        0.02       94.21
         13 |      1,645        5.79      100.00
------------+-----------------------------------
      Total |     28,413      100.00

. tab occ_code, sort

 occupation |      Freq.     Percent        Cum.
------------+-----------------------------------
          3 |     10,974       38.62       38.62
          6 |      4,309       15.17       53.79
          8 |      4,300       15.13       68.92
          1 |      3,008       10.59       79.51
         13 |      1,645        5.79       85.30
          2 |      1,494        5.26       90.56
          4 |      1,323        4.66       95.21
          7 |        571        2.01       97.22
          5 |        438        1.54       98.76
         11 |        194        0.68       99.45
         10 |        144        0.51       99.95
         12 |          7        0.02       99.98
          9 |          6        0.02      100.00
------------+-----------------------------------
      Total |     28,413      100.00

[Rich Goldstein]

do you know that there is a "sort" option to -tab-; does that do what you want? otherwise, see

help recode

[Jean Jacques]

Yes, indeed I knew that but my question is about recoding. How do I recode those 3, 6, 8 (in Nick's example) to 1, 2, 3

[Rich Goldstein]

recode varname (3=1) (6=2) (8=3), gen(varname2)

obviously, you should replace "varname" with the name of your variable and replace "varname2" with whatever name you want for the recoded variable; see

help recode

[Jean Jacques]

Thanks Rich for the answer. But ideally I would like a general way of doing this given that it's I have many variables that I would like to recode.

[Nick Cox]

Here's some technique. You need labmask from the Stata Journal . In general, you can loop over the variables of interest.

webuse nlswork, clear

foreach v in occ_code ind_code { 
    bysort `v' : gen freq = -_N if !missing(`v')
    egen `v'_2 = group(freq `v') 
    capture labmask `v'_2, values(`v') decode
    if _rc labmask `v'_2, values(`v')
    drop freq 
    _crcslbl `v'_2 `v'
}

tab1  *2