As I mentioned earlier in this series of messages, we have written a number of extension commands that tap packages in the R statistical language. I have already discussed SPSSINC RASCH for Rasch models and SPSSINC QUANTREG for quantile regression. Today, I want to introduce
SPSSINC ROBUST REGR and
SPSSINC TOBIT REGR.
SPSSINC ROBUST REGR estimates regression models using an M estimator and is thus more resistant to outliers than is ordinary least squares. As for most of our regression-like R extension commands, it can return the coefficients and residuals as SPSS datasets for further use as well as displaying an SPSS pivot table with the estimation results. Pivot tables from R can also be captured by OMS. (Right click on an instance in the Viewer outline to see the table subtype.) Categorical variables, as determined from the variable measurement levels, are automatically converted to factors. The package includes a dialog box interface that appears on the Analyze>Regression submenu. It is built on the R MASS package.
SPSSINC TOBIT REGR estimates truncated dependent variable models. In its simplest form, tobit analysis is a blend of probit analysis (truncated/not truncated) and regression. This version, built on the R package AER, offers a choice of six error distributions and dependent variable truncation on the left, right, or both sides. The most common case would be left truncation at zero with a Normal distribution. As with SPSSINC ROBUST REGR, the extension package includes both the SPSS Statistics syntax definition and a dialog box interface that appears on the Regression menu. As a former colleague of the late, great James Tobin, the inventor of this method, I am particularly pleased to have this procedure available for SPSS Statistics.
As with the other R-based extension commands, these require both the R and Python Plug-Ins, installation of the relevant R package and the relevant SPSS Statistics extension package.
Both of these extension commands will be posted to SPSS Developer Central soon.
Tomorrow, YARP (Yet Another R Package).
Jon K. Peck
SPSS Inc.
[hidden email] <mailto:
[hidden email]>
(ip) phone 312-651-3435
====================To manage your subscription to SPSSX-L, send a message to
[hidden email] (not to SPSSX-L), with no body text except the
command. To leave the list, send the command
SIGNOFF SPSSX-L
For a list of commands to manage subscriptions, send the command
INFO REFCARD
Locked
