How specifications get lost in translation
Ralf Minkenberg · 2012
The first ideas for ad hoc analyses are born very often without any Programmers or Statisticians involved. Some examples are shown how specifications first are generated and then are developed with different functions involved. It is discussed if the final specifications the programmer get (generally from a Statistician) are still the same as wanted from the author, or are changed on their way to be programming-ready like Chinese whispers. A solution could be to involve the programmer in the specification process as early as possible. But is there a common understanding between specification requestors (key opinion leaders, drug safety, marketing, health economics, ...) and programmers at all? Example processes are described to show different ways specifications can be specified and clarified, and how programmers can handle them. Interaction between Programmers, Statisticians and Medicals are the key to get final analyses which fit the needs of all interested and involved functions.