|
Conference on Quantitative
Social Science Research Using R
(click on the title to download the abstract)
|
|
Achim Zeileis
|
|
Statistical Computing in R: Strategies for Turning Ideas into Software
|
Computing and computational methods play an important role in quantitative
(social science) research. Therefore, both researchers and practitioners typically
also become authors of code/software (for their own applications or simulations),
but more often than not this software is not made available (or not in such a
way that makes it easily reusable by others). Hence, we discuss two questions:
Why should researchers write and publish software? How should such software
be written in order to be efficient and still easy to use? The answer to the first question is that publication of software is in the interest of both, the authors (to
promote their methods/results) and the readers/users (to assure reproducibility
of research). The how is less straightforward because its details depend on the
particular problem at hand. However, it is possible to discuss general strategies for
turning ideas into software. The R system for statistical computing assists these
strategies by providing various language features that facilitate the development
of effective software, e.g., object orientation, functions as first-class objects, and
lexical scope among others. We illustrate how these features have been utilized in
the design of the package sandwich for HC and HAC covariance matrix estimation.
Its application in practice is exemplified by regression modeling of count data.
| |
|
|
|
