rxCovCoef: Covariance and Correlation Matrices for Linear Model Coefficients and Explanatory Variables

Article
2019-07-15

Description

Obtain covariance and correlation matrices for the coefficient estimates within rxLinMod, rxLogit, and rxGlm objects and explanatory variables within rxLinMod and rxLogit objects.

Usage

  rxCovCoef(x)
  rxCorCoef(x)
  rxCovData(x)
  rxCorData(x)

Arguments

`x`

object of class rxLinMod, rxLogit, or rxGlm that satisfies conditions in the Details section.

Details

For rxCovCoef and rxCorCoef, the rxLinMod, rxLogit, or rxGlm object must have been fit with covCoef = TRUE and cube = FALSE. The degrees of freedom must be greater than 0.

For rxCovData and rxCorData, the rxLinMod or rxLogit object must have been fit with an intercept term as well as with covData = TRUE and cube = FALSE.

Value

If p is the number of columns in the model matrix, then

For rxCovCoef a p x p numeric matrix containing the covariances of the model coefficients.

For rxCorCoef a p x p numeric matrix containing the correlations amongst the model coefficients.

For rxCovData a (p - 1) x (p - 1) numeric matrix containing the covariances of the non-intercept terms in the model matrix.

For rxCorData a (p - 1) x (p - 1) numeric matrix containing the correlations amongst the non-intercept terms in the model matrix.

Author(s)

Microsoft Corporation Microsoft Technical Support

Examples


 ## Example 1
 # Get the covariance matrix of the estimated model coefficients
 kyphXdfFileName <- file.path(rxGetOption("sampleDataDir"), "kyphosis.xdf")
 kyphLogitWithCovCoef <-
   rxLogit(Kyphosis ~ Age + Number + Start, data = kyphXdfFileName,
           covCoef = TRUE, reportProgress = 0)
 rxCovCoef(kyphLogitWithCovCoef)

 # Compare results with results from stats::glm function
 data(kyphosis, package = "rpart")
 kyphGlmSummary <-
   summary(glm(Kyphosis ~ Age + Number + Start, data = kyphosis,
               family = binomial()))
 kyphGlmSummary[["cov.scaled"]]


 ## Example 2
 # Get the covariance matrix of the data
 kyphXdfFileName <- file.path(rxGetOption("sampleDataDir"), "kyphosis.xdf")
 kyphLogitWithCovData <-
   rxLogit(Kyphosis ~ Age + Number + Start, data = kyphXdfFileName,
           covData = TRUE, reportProgress = 0)
 rxCovData(kyphLogitWithCovData)

 # Compare results with stats::cov function
 cov(kyphosis[2:4])


 ## Example 3
 # Find the correlation matrices for both the coefficient estimates and the
 # explanatory variables
 rxCorCoef(kyphLogitWithCovCoef)
 rxCorData(kyphLogitWithCovData)

Share via