778 Computational Considerations in Microeconometrics
(see, e.g., Chib, 2004; Geweke, 2005). The preliminary issues are tackled in sections
15.2 and 15.3. Section 15.4 discusses the advantages of flexible functional forms
generated using non- and semiparametric methods. Section 15.5 illustrates some
of the ways in which heterogeneous responses are modeled and how resampling
methods are used in the related econometric inference. Section 15.6 illustrates
the use of simulation assisted estimation in handling endogenous regressors in a
simultaneous nonlinear model of choice and outcomes. Section 15.7 concludes.
15.2 Preliminary
A computer has become the standard tool of the trade for most applied econometri-
cians. It is unimaginable how today’s empirical analyses and advances could have
been performed without one. However, a byproduct of this matrimony between
computing and applied econometrics is today’s necessity that empirical researchers
should know and understand many definitions and specialized computer science
jargon. In this section, we provide a quick introduction to essential concepts, ter-
minology and practical considerations that most researchers come into contact
with while working with a computer.
15.2.1 Programming languages
A computer programming language can be thought of as a set of characters,
along with rules to combine them into words and symbols, that can be used to
express detailed instructions to a computer. Most programming languages used by
economists arehigh-level languages, in the sense that they are not generally specific
to a computer central processing unit (CPU), but instead they are portable across
different computer hardware. These high-level languages can generally be divided
into compiled languages and interpreted languages.
Compiled programming languagesrequire a computer program, calledcompiler,
to translate human-readable text instructions, calledsource code, into alow-level
languagesuch as assembly language or machine language. Mostthird-generation
programming languages(3GL), such as C, C++and Fortran, are compiled languages.
Computer programs written in these languages need to be compiled first, and then
executed.
Interpreted programming languages are implemented by a computer pro-
gram calledinterpreter. The interpreter executes instructions written in these
languages. Popular object-orientedfourth-generation programming languages(4GL),
such as GAUSSTM(Aptech Systems, Inc., Black Diamond, Washington), MATLAB®
(The MathWorksTM, Natick, Massachusetts), R (R Development Core Team,
Vienna, Austria), and Stata®(StataCorp LP, College Station, Texas) are interpreted
languages.
The interested reader is recommended to consult articles by Nerlove (2004) and
Renfro (2004a) for a comprehensive list and history of programming languages
used in econometrics up to 2004.