Review of np software for R

✍ Scribed by Teresa D. Harrison

• Publisher: John Wiley and Sons
• Year: 2008
• Tongue: English
• Weight: 57 KB
• Volume: 23
• Category: Article
• ISSN: 0883-7252
• DOI: 10.1002/jae.1004

✦ Synopsis

The np software written by Tristen Hayfield and Jeff Racine is a program specializing in nonparametric and semiparametric estimation techniques. It is an add-on package for the opensource, object-oriented statistical software R. The program is comprehensive, implementing the most recent nonparametric methods and providing a large degree of flexibility in the estimation specifications. It seamlessly integrates bandwidth selection and estimation of unconditional and conditional density/distribution functions, regression functions (conditional mean, quantile, and modal), and semiparametric index models. After estimation, residuals, gradients, standard errors, and predicted values are easily extracted. The software also utilizes R's graphical capabilities to allow visual interpretation of the estimation. One of the key advantages of the package is its ability to incorporate continuous and discrete data simultaneously. Overall, the package will be extremely useful for applied econometricians interested in employing nonparametric and semiparametric techniques.

2. INSTALLATION AND INITIAL SETUP

Because np is an embedded package, R must be installed prior to installation of np. R and np can be downloaded from http://www.cran.r-project.org. The software runs on all commonly used operating systems (Windows, Unix, Linux, Mac-OS X) and under most current hardware. However, owing to the computational intensity of nonparametric estimation, ample RAM and a fast CPU are desirable. This review was performed using a 1.86 GHz Pentium M laptop with 2 GB of RAM and running 2002 Windows XP, Service Pack 2. I used version 2.5 and 0.13-1 of R and np, respectively.