But you have done your due diligence here. Let me emphasize that it is never safe to use -force- unless you have carefully inspected the problematic cases and you know for sure that you are not going to lose vital information (as opposed to meaningless excess precision). But in this case, if you specify a format that gives you as many decimal places in the string version as you need for practical purposes (maybe %3.2f will do the trick), you will have a sufficiently accurate representation of the information. It worked well with most variables but v116 was shown 'cannot be converted reversibly no replace'. The command I used was 'tostring v110-v124, replace'. I'm glad that your first reaction was not to use it: -force- can be dangerous. Since some of them are string and others are numeric, I used tostring command to match their type. This is one of those rare situations where it is appropriate to use the -force- option. It will remove the nonnumeric characters, leaving you with the numeric. There basic function can be described as follows: destring is for variables that contain both numeric and nonnumeric characters (e.g., percentages together with a '' sign). So Stata is telling you that no matter how many places it tries to give you in the string, the information will not be exactly correct, and, in particular, if you try to convert it back to numeric you may not get back what you started with. The tostring, destring, encode and decode commands are available for these more complex changes. The decimal numbers 15.39 and 84.9 have no exact finite binary representation, just as the number 1/3 has no exact finite decimal representation. All floating point numbers in Stata are represented using a binary floating-point format of finite precision. The reason Stata is balking at v116 is because of the values 15.39 and 84.9. In Stata you need to convert this string variable to a date variable.

You can't use -destring- on a variable that isn't already a string: and it does the reverse of what you want.

Tostring stata