Ytukyg

I have string variables in my data that contain fractions and non-integers.

For example:

3 2/3 

8 1/2

 1.65

6 1/4

0.235

Stata cannot destring these variables with the destring command.

Is there anything I can do to convert my string variables to numeric so I can use them for analysis?

Below code, while extending to cases like " 3 2/3" or "- 6 1/4" (notice the spaces), might also be useful for saving some typings.

replace var1 = subinstr(var1, "- ", "-",.)

split var1, generate(x) destring force



egen y = ends(var1) if strmatch(var1,"*/*"), last

split y, destring force p("/")



replace x1 = cond(x1<0, x1-y1/y2, cond(x1!=.,x1+y1/y2, y1/y2)) if y1!=.

keep var1 x1 

If all values in your string variables look the same as in your example, you can do the following:

clear



input str6 var1

"3 2/3"

"8 1/2"

"6/8" 

"-2/3"

"1.65"

"-6 1/4"

"-0.235"

end



replace var1 = subinstr(var1," ",":", .)

replace var1 = "0:" + var1 if !strmatch(var1, "*:*") & !strmatch(var1, "-*") & strmatch(var1, "*/*")

replace var1 = "-0:" + substr(var1, 2, .) if !strmatch(var1, "*:*") & strmatch(var1, "-*") & strmatch(var1, "*/*") 

replace var1 = subinstr(var1,":"," ", .)



generate double var2 = real(var1) if !strmatch(var1, "*/*")



replace var2 = real(substr(var1, 1, strpos(var1, " ") - 1)) + ///

               ( real(substr(var1, strpos(var1, " "), strpos(var1, "/") - strpos(var1, " "))) / ///

               real(substr(var1, strpos(var1, "/") + 1, .)) ) ///

               if strmatch(var1, "*/*") & !strmatch(var1, "-*")



replace var2 = - ( abs(real(substr(var1, 1, strpos(var1, " ") - 1))) + ///

               ( real(substr(var1, strpos(var1, " "), strpos(var1, "/") - strpos(var1, " "))) / ///

               real(substr(var1, strpos(var1, "/") + 1, .)) ) ) ///

               if strmatch(var1, "*/*") & strmatch(var1, "-*")

For more information on the use of each function, type help followed by its
name: real(), strmatch(), substr(), subinstr(), strpos(), abs().

EDIT:

The aforementioned approach relies on string manipulation using Stata's string functions and hence is lengthier than the solution provided in the other answer.

Nevertheless, you can shorten the code a bit and improve readability as follows:

local m1 !strmatch(var1, "* *")

local m2 strmatch(var1, "-*") & strmatch(var1, "*/*")



replace var1 = cond(`m1' & !`m2', "0 " + var1, ///

               cond(`m1' & `m2', "-0 " + substr(var1, 2, .), var1 + "" )  )



local e1  real(substr(var1, 1, strpos(var1, " ") - 1))

local e2 (real(substr(var1, strpos(var1, " "), strpos(var1, "/") - strpos(var1, " "))) / ///

          real(substr(var1, strpos(var1, "/") + 1, .)) )



generate var2 = cond(!strmatch(var1, "*/*"), real(var1), ///

                cond(!`m2', `e1' + `e2', - ( abs(`e1') + `e2' ) ) )

In both cases, you will get:

list, separator(0)



     +---------------------+

     |   var1         var2 |

     |---------------------|

  1. |  3 2/3    3.6666667 |

  2. |  8 1/2          8.5 |

  3. |  0 6/8          .75 |

  4. | -0 2/3   -.66666667 |

  5. |   1.65         1.65 |

  6. | -6 1/4        -6.25 |

  7. | -0.235        -.235 |

     +---------------------+