Stata tip 148: Searching for words within strings

Abstract

Searching for particular text within strings is a common data management problem. One frequent context is whenever various possible answers to a question are bundled together in values of a string variable. Suppose people are asked which sports they enjoy or something more interesting, like which statistical software they use routinely. To keep the matter simple, we will first imagine just lists of one or more numbers that are concise codes for distinct answers, say, "42" for "cricket" or "1" for "Stata". Nonnumeric codes will also be considered in due course. For more on handling such questions, sometimes called multiple response, see Cox and Kohler (2003) or Jann (2005).