pcreg: Principal Component and Partial Least Squares Regression
In bestglm: Best Subset GLM and Regression Utilities

Description Usage Arguments Value Author(s) See Also Examples

Regression using the principal components or latent variables as inputs. The best model is selected using components 1, 2, ..., r, where r, the number of components to use is determined by the AIC or BIC.

1	pcreg(Xy, scale = TRUE, method = c("PC", "LV"), ic = c("BIC", "AIC"))

`Xy`	dataframe with variable names in columns
`scale`	Whether or not to scale. Default is TRUE.
`method`	either principal components, "PC", or partial least squares latent variables, "LV"
`ic`	"BIC" or "AIC"

An S3 class list "pcreg" with components

`lmfit`	lm model
`PLSFit`	column sd
`Z`	matrix of principal components or latent vector
`method`	'pcr' or 'pls'

A. I. McLeod

predict.pcreg, summary.pcreg, plot.pcreg, fitted.pcreg, residuals.pcreg

1 2	pcreg(mcdonald, scale=TRUE, method="PC") pcreg(mcdonald, scale=TRUE, method="LV")

Loading required package: leaps

Call:
lm(formula = MORT ~ ., data = Zy[, c(1:mIC, p + 1)])

Coefficients:
(Intercept)          PC1          PC2          PC3          PC4          PC5  
   940.3584      15.5877      -3.2915     -19.8282      -2.7007      -0.7183  
        PC6          PC7  
    21.0846      17.3389  


Call:
lm(formula = MORT ~ ., data = Zy[, c(1:mIC, p + 1)])

Coefficients:
(Intercept)          LV1          LV2          LV3  
     940.36        23.60        17.40        14.91