Is there an elegant way to assign midpoints to data?

[Sofiya Volvakova]

Hi Statalist,

I need to assign midpoints to each band of the household's gross total income. I understand how midpoints work (max - min)/2. However, I am not sure whether doing it manually is the most elegant solution. Is there a command in Stata that can do it for me ?

Here is the excerpt of the data:

DV: Gross |
household income |
Freq. Percent Cum.
------------------------+-----------------------------------
Less than �520 | 17 0.14 0.14
�520 less than �1040 | 17 0.14 0.29
�1040 less than �1560 | 26 0.22 0.51
�1560 less than �2080 | 47 0.40 0.91
�2080 less than �2600 | 61 0.52 1.43
�2600 Less than �3120 | 40 0.34 1.77
�3120 less than �3640 | 41 0.35 2.12
�3640 less than �4160 | 61 0.52 2.64
�4160 less than �4680 | 67 0.57 3.21
�4680 less than �5200 | 98 0.84 4.05
�5200 Less than �6240 | 326 2.78 6.83
�6240 less than �7280 | 340 2.90 9.73
�7280 less than �8320 | 379 3.23 12.96
�8320 less than �9360 | 355 3.03 15.98
�9360 less than �10400 | 455 3.88 19.86
�10400 less than �11440 | 452 3.85 23.71
�11440 less than �12480 | 400 3.41 27.12
�12480 less than �13520 | 442 3.77 30.89
�13520 less than �14560 | 342 2.92 33.81
�14560 less than �15600 | 408 3.48 37.29
�15600 less than �16640 | 328 2.80 40.08
�16640 less than �17680 | 283 2.41 42.49
�17680 less than �18720 | 271 2.31 44.80
�18720 less than �19760 | 319 2.72 47.52
�19760 less than �20800 | 365 3.11 50.64
�20800 less than �23400 | 617 5.26 55.89
�23400 less than �26000 | 616 5.25 61.15
�26000 less than �28600 | 513 4.37 65.52
�28600 less than �31200 | 523 4.46 69.98
�31200 less than �33800 | 559 4.77 74.74
�33800 less than �36400 | 466 3.97 78.71
�36,400 or more | 2,497 21.29 100.00
------------------------+-----------------------------------
Total | 11,731 100.00

[Tiago Pereira]

If possible, can you provide us with some data via -dataex-?

[Sofiya Volvakova]

If possible, can you provide us with some data via -dataex-?

Unfortunately, I cannot share the dataset and copying/pasting the variable into the new dataset converted it from string to numeric. I don't need the exact code for Stata. I just want to know what would be the appropriate Stata syntax based on the data excerpt I provided above. Thanks, Tiago

[Andrew Musau]

Tiago Pereira is right that we need a data example. It appears that your start-points and end-points are in the value labels, so you can use decode combined with other string functions and generate the wanted variable.

[Sofiya Volvakova]

Tiago Pereira is right that we need a data example. It appears that your start-points and end-points are in the value labels, so you can use decode combined with other string functions and generate the wanted variable.

----------------------- copy starting from the next line -----------------------

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte W1inc1est
 26
 32
 26
-99
 20
 21
 26
 26
 32
-92
-92
 27
 15
 30
 15
 12
 32
 -1
 12
 17
 27
-92
 31
 17
 16
 -1
 -1
 28
 14
 32
 31
 11
-92
 17
 -1
 -1
 -1
 21
 -1
 28
 21
 12
 -1
 16
 -1
-92
 32
 32
 14
 27
 11
 14
 23
 32
 14
-92
-92
-92
 29
 26
 21
 -1
 26
 29
-92
 16
  9
-99
 16
 28
 32
 23
 17
 31
 28
 32
 -1
 32
 24
 17
 18
-92
-92
 31
 32
 32
 32
 31
-92
 14
 18
 18
 20
 14
 19
 25
 16
 32
 -1
 26
end
label values W1inc1est W1inc1est
label def W1inc1est -99 "MP not interviewed", modify
label def W1inc1est -92 "Refused", modify
label def W1inc1est -1 "Don't know", modify
label def W1inc1est 9 "�4160 less than �4680", modify
label def W1inc1est 11 "�5200 Less than �6240", modify
label def W1inc1est 12 "�6240 less than �7280", modify
label def W1inc1est 14 "�8320 less than �9360", modify
label def W1inc1est 15 "�9360 less than �10400", modify
label def W1inc1est 16 "�10400 less than �11440", modify
label def W1inc1est 17 "�11440 less than �12480", modify
label def W1inc1est 18 "�12480 less than �13520", modify
label def W1inc1est 19 "�13520 less than �14560", modify
label def W1inc1est 20 "�14560 less than �15600", modify
label def W1inc1est 21 "�15600 less than �16640", modify
label def W1inc1est 23 "�17680 less than �18720", modify
label def W1inc1est 24 "�18720 less than �19760", modify
label def W1inc1est 25 "�19760 less than �20800", modify
label def W1inc1est 26 "�20800 less than �23400", modify
label def W1inc1est 27 "�23400 less than �26000", modify
label def W1inc1est 28 "�26000 less than �28600", modify
label def W1inc1est 29 "�28600 less than �31200", modify
label def W1inc1est 30 "�31200 less than �33800", modify
label def W1inc1est 31 "�33800 less than �36400", modify
label def W1inc1est 32 "�36,400 or more", modify

------------------ copy up to and including the previous line ------------------

Listed 100 out of 15770 observations
Use the count() option to list more

[Nick Cox]

The best recipe for a midpoint is (max + min) / 2 as (max - min) / 2 is half the range, a different beast.

[Sofiya Volvakova]

The best recipe for a midpoint is (max + min) / 2 as (max - min) / 2 is half the range, a different beast.

Thank you for correcting, Nick! Is there a command for Stata to tell it to derive a midpoint for those income bands? Or will I have to do it manually with the "recode" function, e.g recode W1inc1est (1=250) (2=500), etc.?

[Nick Cox]

At a quick glance your bins vary in width, so you will waste time trying to code them any way but directly.

[Andrew Musau]

There are no mid-points for intervals defined as \(amount>X\). So I just return the amount below.

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte W1inc1est
 26
 32
 26
-99
 20
 21
 26
 26
 32
-92
-92
 27
 15
 30
 15
 12
 32
 -1
 12
 17
 27
-92
 31
 17
 16
 -1
 -1
 28
 14
 32
 31
 11
-92
 17
 -1
 -1
 -1
 21
 -1
 28
 21
 12
 -1
 16
 -1
-92
 32
 32
 14
 27
 11
 14
 23
 32
 14
-92
-92
-92
 29
 26
 21
 -1
 26
 29
-92
 16
  9
-99
 16
 28
 32
 23
 17
 31
 28
 32
 -1
 32
 24
 17
 18
-92
-92
 31
 32
 32
 32
 31
-92
 14
 18
 18
 20
 14
 19
 25
 16
 32
 -1
 26
end
label values W1inc1est W1inc1est
label def W1inc1est -99 "MP not interviewed", modify
label def W1inc1est -92 "Refused", modify
label def W1inc1est -1 "Don't know", modify
label def W1inc1est 9 "�4160 less than �4680", modify
label def W1inc1est 11 "�5200 Less than �6240", modify
label def W1inc1est 12 "�6240 less than �7280", modify
label def W1inc1est 14 "�8320 less than �9360", modify
label def W1inc1est 15 "�9360 less than �10400", modify
label def W1inc1est 16 "�10400 less than �11440", modify
label def W1inc1est 17 "�11440 less than �12480", modify
label def W1inc1est 18 "�12480 less than �13520", modify
label def W1inc1est 19 "�13520 less than �14560", modify
label def W1inc1est 20 "�14560 less than �15600", modify
label def W1inc1est 21 "�15600 less than �16640", modify
label def W1inc1est 23 "�17680 less than �18720", modify
label def W1inc1est 24 "�18720 less than �19760", modify
label def W1inc1est 25 "�19760 less than �20800", modify
label def W1inc1est 26 "�20800 less than �23400", modify
label def W1inc1est 27 "�23400 less than �26000", modify
label def W1inc1est 28 "�26000 less than �28600", modify
label def W1inc1est 29 "�28600 less than �31200", modify
label def W1inc1est 30 "�31200 less than �33800", modify
label def W1inc1est 31 "�33800 less than �36400", modify
label def W1inc1est 32 "�36,400 or more", modify

decode W1inc1est, gen(newvar)
gen tosplit= trim(itrim(ustrregexra(newvar, "[^\d,]", " ")))
split tosplit, g(limit) ignore(",") destring
gen wanted= cond(missing(limit2), limit1, (limit1+limit2)/2)
drop newvar tosplit

Res.:

. l, sep(0)

     +----------------------------------------------------+
     |               W1inc1est   limit1   limit2   wanted |
     |----------------------------------------------------|
  1. | �20800 less than �23400    20800    23400    22100 |
  2. |         �36,400 or more    36400        .    36400 |
  3. | �20800 less than �23400    20800    23400    22100 |
  4. |      MP not interviewed        .        .        . |
  5. | �14560 less than �15600    14560    15600    15080 |
  6. | �15600 less than �16640    15600    16640    16120 |
  7. | �20800 less than �23400    20800    23400    22100 |
  8. | �20800 less than �23400    20800    23400    22100 |
  9. |         �36,400 or more    36400        .    36400 |
 10. |                 Refused        .        .        . |
 11. |                 Refused        .        .        . |
 12. | �23400 less than �26000    23400    26000    24700 |
 13. |  �9360 less than �10400     9360    10400     9880 |
 14. | �31200 less than �33800    31200    33800    32500 |
 15. |  �9360 less than �10400     9360    10400     9880 |
 16. |   �6240 less than �7280     6240     7280     6760 |
 17. |         �36,400 or more    36400        .    36400 |
 18. |              Don't know        .        .        . |
 19. |   �6240 less than �7280     6240     7280     6760 |
 20. | �11440 less than �12480    11440    12480    11960 |
 21. | �23400 less than �26000    23400    26000    24700 |
 22. |                 Refused        .        .        . |
 23. | �33800 less than �36400    33800    36400    35100 |
 24. | �11440 less than �12480    11440    12480    11960 |
 25. | �10400 less than �11440    10400    11440    10920 |
 26. |              Don't know        .        .        . |
 27. |              Don't know        .        .        . |
 28. | �26000 less than �28600    26000    28600    27300 |
 29. |   �8320 less than �9360     8320     9360     8840 |
 30. |         �36,400 or more    36400        .    36400 |
 31. | �33800 less than �36400    33800    36400    35100 |
 32. |   �5200 Less than �6240     5200     6240     5720 |
 33. |                 Refused        .        .        . |
 34. | �11440 less than �12480    11440    12480    11960 |
 35. |              Don't know        .        .        . |
 36. |              Don't know        .        .        . |
 37. |              Don't know        .        .        . |
 38. | �15600 less than �16640    15600    16640    16120 |
 39. |              Don't know        .        .        . |
 40. | �26000 less than �28600    26000    28600    27300 |
 41. | �15600 less than �16640    15600    16640    16120 |
 42. |   �6240 less than �7280     6240     7280     6760 |
 43. |              Don't know        .        .        . |
 44. | �10400 less than �11440    10400    11440    10920 |
 45. |              Don't know        .        .        . |
 46. |                 Refused        .        .        . |
 47. |         �36,400 or more    36400        .    36400 |
 48. |         �36,400 or more    36400        .    36400 |
 49. |   �8320 less than �9360     8320     9360     8840 |
 50. | �23400 less than �26000    23400    26000    24700 |
 51. |   �5200 Less than �6240     5200     6240     5720 |
 52. |   �8320 less than �9360     8320     9360     8840 |
 53. | �17680 less than �18720    17680    18720    18200 |
 54. |         �36,400 or more    36400        .    36400 |
 55. |   �8320 less than �9360     8320     9360     8840 |
 56. |                 Refused        .        .        . |
 57. |                 Refused        .        .        . |
 58. |                 Refused        .        .        . |
 59. | �28600 less than �31200    28600    31200    29900 |
 60. | �20800 less than �23400    20800    23400    22100 |
 61. | �15600 less than �16640    15600    16640    16120 |
 62. |              Don't know        .        .        . |
 63. | �20800 less than �23400    20800    23400    22100 |
 64. | �28600 less than �31200    28600    31200    29900 |
 65. |                 Refused        .        .        . |
 66. | �10400 less than �11440    10400    11440    10920 |
 67. |   �4160 less than �4680     4160     4680     4420 |
 68. |      MP not interviewed        .        .        . |
 69. | �10400 less than �11440    10400    11440    10920 |
 70. | �26000 less than �28600    26000    28600    27300 |
 71. |         �36,400 or more    36400        .    36400 |
 72. | �17680 less than �18720    17680    18720    18200 |
 73. | �11440 less than �12480    11440    12480    11960 |
 74. | �33800 less than �36400    33800    36400    35100 |
 75. | �26000 less than �28600    26000    28600    27300 |
 76. |         �36,400 or more    36400        .    36400 |
 77. |              Don't know        .        .        . |
 78. |         �36,400 or more    36400        .    36400 |
 79. | �18720 less than �19760    18720    19760    19240 |
 80. | �11440 less than �12480    11440    12480    11960 |
 81. | �12480 less than �13520    12480    13520    13000 |
 82. |                 Refused        .        .        . |
 83. |                 Refused        .        .        . |
 84. | �33800 less than �36400    33800    36400    35100 |
 85. |         �36,400 or more    36400        .    36400 |
 86. |         �36,400 or more    36400        .    36400 |
 87. |         �36,400 or more    36400        .    36400 |
 88. | �33800 less than �36400    33800    36400    35100 |
 89. |                 Refused        .        .        . |
 90. |   �8320 less than �9360     8320     9360     8840 |
 91. | �12480 less than �13520    12480    13520    13000 |
 92. | �12480 less than �13520    12480    13520    13000 |
 93. | �14560 less than �15600    14560    15600    15080 |
 94. |   �8320 less than �9360     8320     9360     8840 |
 95. | �13520 less than �14560    13520    14560    14040 |
 96. | �19760 less than �20800    19760    20800    20280 |
 97. | �10400 less than �11440    10400    11440    10920 |
 98. |         �36,400 or more    36400        .    36400 |
 99. |              Don't know        .        .        . |
100. | �20800 less than �23400    20800    23400    22100 |
     +----------------------------------------------------+

.

[Sofiya Volvakova]

There are no mid-points for intervals defined as \(amount>X\). So I just return the amount below.

* Example generated by -dataex-. To install: ssc install dataex
clear
input byte W1inc1est
26
32
26
-99
20
21
26
26
32
-92
-92
27
15
30
15
12
32
-1
12
17
27
-92
31
17
16
-1
-1
28
14
32
31
11
-92
17
-1
-1
-1
21
-1
28
21
12
-1
16
-1
-92
32
32
14
27
11
14
23
32
14
-92
-92
-92
29
26
21
-1
26
29
-92
16
9
-99
16
28
32
23
17
31
28
32
-1
32
24
17
18
-92
-92
31
32
32
32
31
-92
14
18
18
20
14
19
25
16
32
-1
26
end
label values W1inc1est W1inc1est
label def W1inc1est -99 "MP not interviewed", modify
label def W1inc1est -92 "Refused", modify
label def W1inc1est -1 "Don't know", modify
label def W1inc1est 9 "�4160 less than �4680", modify
label def W1inc1est 11 "�5200 Less than �6240", modify
label def W1inc1est 12 "�6240 less than �7280", modify
label def W1inc1est 14 "�8320 less than �9360", modify
label def W1inc1est 15 "�9360 less than �10400", modify
label def W1inc1est 16 "�10400 less than �11440", modify
label def W1inc1est 17 "�11440 less than �12480", modify
label def W1inc1est 18 "�12480 less than �13520", modify
label def W1inc1est 19 "�13520 less than �14560", modify
label def W1inc1est 20 "�14560 less than �15600", modify
label def W1inc1est 21 "�15600 less than �16640", modify
label def W1inc1est 23 "�17680 less than �18720", modify
label def W1inc1est 24 "�18720 less than �19760", modify
label def W1inc1est 25 "�19760 less than �20800", modify
label def W1inc1est 26 "�20800 less than �23400", modify
label def W1inc1est 27 "�23400 less than �26000", modify
label def W1inc1est 28 "�26000 less than �28600", modify
label def W1inc1est 29 "�28600 less than �31200", modify
label def W1inc1est 30 "�31200 less than �33800", modify
label def W1inc1est 31 "�33800 less than �36400", modify
label def W1inc1est 32 "�36,400 or more", modify

decode W1inc1est, gen(newvar)
gen tosplit= trim(itrim(ustrregexra(newvar, "[^\d,]", " ")))
split tosplit, g(limit) ignore(",") destring
gen wanted= cond(missing(limit2), limit1, (limit1+limit2)/2)
drop newvar tosplit

Res.:

. l, sep(0)

+----------------------------------------------------+
| W1inc1est limit1 limit2 wanted |
|----------------------------------------------------|
1. | �20800 less than �23400 20800 23400 22100 |
2. | �36,400 or more 36400 . 36400 |
3. | �20800 less than �23400 20800 23400 22100 |
4. | MP not interviewed . . . |
5. | �14560 less than �15600 14560 15600 15080 |
6. | �15600 less than �16640 15600 16640 16120 |
7. | �20800 less than �23400 20800 23400 22100 |
8. | �20800 less than �23400 20800 23400 22100 |
9. | �36,400 or more 36400 . 36400 |
10. | Refused . . . |
11. | Refused . . . |
12. | �23400 less than �26000 23400 26000 24700 |
13. | �9360 less than �10400 9360 10400 9880 |
14. | �31200 less than �33800 31200 33800 32500 |
15. | �9360 less than �10400 9360 10400 9880 |
16. | �6240 less than �7280 6240 7280 6760 |
17. | �36,400 or more 36400 . 36400 |
18. | Don't know . . . |
19. | �6240 less than �7280 6240 7280 6760 |
20. | �11440 less than �12480 11440 12480 11960 |
21. | �23400 less than �26000 23400 26000 24700 |
22. | Refused . . . |
23. | �33800 less than �36400 33800 36400 35100 |
24. | �11440 less than �12480 11440 12480 11960 |
25. | �10400 less than �11440 10400 11440 10920 |
26. | Don't know . . . |
27. | Don't know . . . |
28. | �26000 less than �28600 26000 28600 27300 |
29. | �8320 less than �9360 8320 9360 8840 |
30. | �36,400 or more 36400 . 36400 |
31. | �33800 less than �36400 33800 36400 35100 |
32. | �5200 Less than �6240 5200 6240 5720 |
33. | Refused . . . |
34. | �11440 less than �12480 11440 12480 11960 |
35. | Don't know . . . |
36. | Don't know . . . |
37. | Don't know . . . |
38. | �15600 less than �16640 15600 16640 16120 |
39. | Don't know . . . |
40. | �26000 less than �28600 26000 28600 27300 |
41. | �15600 less than �16640 15600 16640 16120 |
42. | �6240 less than �7280 6240 7280 6760 |
43. | Don't know . . . |
44. | �10400 less than �11440 10400 11440 10920 |
45. | Don't know . . . |
46. | Refused . . . |
47. | �36,400 or more 36400 . 36400 |
48. | �36,400 or more 36400 . 36400 |
49. | �8320 less than �9360 8320 9360 8840 |
50. | �23400 less than �26000 23400 26000 24700 |
51. | �5200 Less than �6240 5200 6240 5720 |
52. | �8320 less than �9360 8320 9360 8840 |
53. | �17680 less than �18720 17680 18720 18200 |
54. | �36,400 or more 36400 . 36400 |
55. | �8320 less than �9360 8320 9360 8840 |
56. | Refused . . . |
57. | Refused . . . |
58. | Refused . . . |
59. | �28600 less than �31200 28600 31200 29900 |
60. | �20800 less than �23400 20800 23400 22100 |
61. | �15600 less than �16640 15600 16640 16120 |
62. | Don't know . . . |
63. | �20800 less than �23400 20800 23400 22100 |
64. | �28600 less than �31200 28600 31200 29900 |
65. | Refused . . . |
66. | �10400 less than �11440 10400 11440 10920 |
67. | �4160 less than �4680 4160 4680 4420 |
68. | MP not interviewed . . . |
69. | �10400 less than �11440 10400 11440 10920 |
70. | �26000 less than �28600 26000 28600 27300 |
71. | �36,400 or more 36400 . 36400 |
72. | �17680 less than �18720 17680 18720 18200 |
73. | �11440 less than �12480 11440 12480 11960 |
74. | �33800 less than �36400 33800 36400 35100 |
75. | �26000 less than �28600 26000 28600 27300 |
76. | �36,400 or more 36400 . 36400 |
77. | Don't know . . . |
78. | �36,400 or more 36400 . 36400 |
79. | �18720 less than �19760 18720 19760 19240 |
80. | �11440 less than �12480 11440 12480 11960 |
81. | �12480 less than �13520 12480 13520 13000 |
82. | Refused . . . |
83. | Refused . . . |
84. | �33800 less than �36400 33800 36400 35100 |
85. | �36,400 or more 36400 . 36400 |
86. | �36,400 or more 36400 . 36400 |
87. | �36,400 or more 36400 . 36400 |
88. | �33800 less than �36400 33800 36400 35100 |
89. | Refused . . . |
90. | �8320 less than �9360 8320 9360 8840 |
91. | �12480 less than �13520 12480 13520 13000 |
92. | �12480 less than �13520 12480 13520 13000 |
93. | �14560 less than �15600 14560 15600 15080 |
94. | �8320 less than �9360 8320 9360 8840 |
95. | �13520 less than �14560 13520 14560 14040 |
96. | �19760 less than �20800 19760 20800 20280 |
97. | �10400 less than �11440 10400 11440 10920 |
98. | �36,400 or more 36400 . 36400 |
99. | Don't know . . . |
100. | �20800 less than �23400 20800 23400 22100 |
+----------------------------------------------------+

.

Hi again! Thanks so much for your help, I definitely did not know such commands so will make sure to go through them thoroughly . Thanks a mil, you really helped!

[Griffin JM Bur]

I came across this thread while attempting to do something similar, to assign numeric values to an ordinal income variable which has the first k integers as existing values but scaled, variable-range bands as the value label. Many American social research data-sets have this for their income label, so this seems like a useful general tool to me. Below are the data (from the General Social Survey 1994), and after that is my code.

* Example generated by -dataex-. For more info, type help dataex
clear
input float rincom91
22
11
12
 9
14
 9
12
12
22
18
16
21
18
18
 9
 9
20
15
19
16
10
21
18
22
21
14
18
 2
14
12
18
10
12
15
16
22
22
15
10
16
22
 9
 8
 5
21
21
21
21
 8
 7
22
20
22
 2
22
19
20
21
19
 2
11
18
12
 3
16
13
22
 3
 9
 4
22
15
12
14
10
13
18
13
19
 4
21
15
18
17
17
19
 8
15
18
16
 8
16
12
12
20
18
19
15
21
14
end
label values rincom91 rincom91
label def rincom91 2 "$1000-2999", modify
label def rincom91 3 "$3000-3999", modify
label def rincom91 4 "$4000-4999", modify
label def rincom91 5 "$5000-5999", modify
label def rincom91 7 "$7000-7999", modify
label def rincom91 8 "$8000-9999", modify
label def rincom91 9 "$10000-12499", modify
label def rincom91 10 "$12500-14999", modify
label def rincom91 11 "$15000-17499", modify
label def rincom91 12 "$17500-19999", modify
label def rincom91 13 "$20000-22499", modify
label def rincom91 14 "$22500-24999", modify
label def rincom91 15 "$25000-29999", modify
label def rincom91 16 "$30000-34999", modify
label def rincom91 17 "$35000-39999", modify
label def rincom91 18 "$40000-49999", modify
label def rincom91 19 "$50000-59999", modify
label def rincom91 20 "$60000-74999", modify
label def rincom91 21 "$75000+", modify
label def rincom91 22 "refused", modify

The following code will need some modifications depending on the data-set, of course; sorry for posting an answer that works specifically for my data, but hopefully it's still useful (you might want to use the word() function).

gen realinc = .
        
        lab def rincom91 1 "$0-1000", modify
        
        * Here's a fast way to take value labels and do midpoint imputation 
        * using them, inspired to some extent by this answer on Stackexchange: 
            * https://archive.ph/HxE2G
        
        levelsof rincom91, local(levs)
            * This stores all possible values of the variable in a local 
            * called "levs"
        local inclab : value label rincom91
            * We also put the value label itself into a local, using Stata's
            * extended macro functions, which let us use the syntax above (for
            * more on this, see p. 4 of this and on: 
                * https://www.stata.com/manuals13/pmacro.pdf)

        foreach k of local levs {
            local strlab : label `inclab' `k'
                * What's happening here? We access the string label associated
                    * with numeric value k, for k in the set of all possible
                    * values of income
            local n2 = strrpos("`strlab'", "-") - 2
                * Now, I want to turn that string into two different numbers
                * but the problem is that the numbers are of variable length
                * so I just say "tell me at what position in the string the 
                * hyphen first occurs" and store that as n2. I subtract two 
                * for reasons mentioned below
            local lb = substr("`strlab'", 2, `n2')
                * Now, I get the lower bound of the interval by taking the sub-
                * string from the second position of the string label for a 
                * length of n2. Now we see why I took away two above in defining
                * n2: I didn't want to include the hyphen itself, and I was 
                * starting the string after the first character, a dollar sign
            local ub = substr("`strlab'", `n2'+3, strlen("`strlab'"))
                * Now I get the ub. I add three to n2 to start the second string
                * after the hyphen and then go until the end of the string label
                * using the length of the string
            replace realinc = round((real("`lb'") +  real("`ub'"))/2, 1)  ///
                if rincom91 == `k'
                * Finally, I replace income with the average of my two values
        }