colldx: Collinearity Diagnostics for Nonlinear Models Using the...
In rpkyle/epimisc: Miscellaneous R functions for basic epidemiologic analysis.

Description Usage Arguments Details Author(s) References Examples

The colldx function implements the Belsley, Kuh, and Welsch approach to diagnosing collinearity.

1	colldx(mod, digits)

`mod`	a model object for one of several supported model classes.
`digits`	minimal number of significant digits, see `print.default`.

The colldx function is the R version of a SAS macro originally titled collingenmodv9c. Beau Bruce adapted the colldiag function from the perturb package to compute condition indices as in Matthew Zack's SAS macro.

The macro (and this R function) implement a method suggested by Belsley, Kuh, and Welsch in Regression Diagnostics: Identifying Influential Data and Sources of Collinearity (1980).

Several model classes are currently supported (though have not been extensively tested): glm, coxph, geeglm, geese, gee, lmerMod, glmerMod, glmmML, and yagsResult.

Written by Beau Bruce, modified for compatibility with various model classes by Ryan Patrick Kyle <ryan.kyle@mail.mcgill.ca>.

Belsley DA. A Guide to Using the Collinearity Diagnostics. Computer Science in Economics and Management 2014; 4:33-50.

# As used by perturb/colldiag, example of SAS collinearity diagnostics from
# https://stats.idre.ucla.edu/sas/webbooks/reg/chapter2/regressionwith-saschapter-2-regression-diagnostics/
# 2.4 Tests for Collinearity
library(foreign)
elemapi <- read.dta("https://stats.idre.ucla.edu/stat/data/elemapi2.dta")
m2 <- lm(api00 ~ acs_k3+avg_ed+grad_sch+col_grad+some_col, data=elemapi)
colldx(m2, digits=3)