Psicometría avanzada con R a partir de datos de personalidad en myPersonality
Paulo Villegas (Telefónica Investigación y Desarrollo)
Pedro Concejero (Telefónica Investigación y Desarrollo)
Miguel Angel Castellanos (Universidad Complutense de Madrid)
Psychometrics is the discipline devoted to psychological
measurement. This is a highly specialized field, and somewhat confined
within the frontiers of psychological testing, but its most modern
models have many applications. These are called latent trait models,
or also known (within those frontiers) as Item Response Theory, or
IRT (http://en.wikipedia.org/wiki/Item_response_theory). A main
difficulty for applying these models has been, until very recently,
the limited availability of software to compute them, usually obscure
commercial solutions (we mean only known to very specific academic and
research communities) with lots of inconveniences to integrate with
usual statistical software.
The R statistical software framework has promoted a small revolution
within this field, allowing many researchers around the world to
contribute libraries that estimate a plethora of these models, with an
efficiency not known till now. Our purpose with this paper is to
present the application of advanced IRT models, as implemented by R
packages
(http://cran.r-project.org/web/views/Psychometrics.html), to a
popular Facebook application devoted to personality measurement,
called myPersonality
(http://www.facebook.com/apps/application.php?id=2490151219). A
dataset of nearly one million user-filled complete tests
(http://www.mypersonality.org/wiki/doku.php) is used to compute
the models within R and analyze their results. Applications of these
models for subjective measurement problems, like preference modeling,
are also outlined.
organizacion@usar.org.es