Erehweb’s Blog has a fascinating discussion that challenges us, as statistical programmers and analysts, to think hard about the tools we use and the ways in which we use them.
The author makes a point about the success of the file-sharing product called Dropbox. He advocates that this product does something really well and in a straight-forward manner and challenges statistical programming languages to follow that same paradigm. From the many vehement comments, it is clear that he has raised a controversial point.
Are the statistical programming languages we use packed with features that aren't truly needed? Do they require hard-to-read code when sometimes all our customers want is something simple?
I've thought about this for a while now. To my surprise, I've concluded that complexity is okay! I work in a complicated business and a fair amount of detail and nuance are needed to fully understand the statistical summaries and reports I provide.
For example, in a recent project, I was thinking about providing a "traffic light" type of report that showed whether some historical work was good (green) or bad (red) or ambivalent (yellow). After some thought and interchange with clients, I decided that a more detailed report would be required.
Simple isn't always best.