!! Compute Standardized Regression Coefficients

Source: R/coef_standardized.R

Compute the standardized regression coefficients (beta) from an object of class lm).

coef_standardized(obj)

# S3 method for lm_beta
print(x, ..., digits = 3)

# S3 method for lm_beta
summary(object, ..., digits = 3)

# S3 method for lm_beta_summary
print(x, ..., digits = 3)

Arguments

obj

(lm object) A result of function lm().
x

lm_beta object.
...

further parameters to print method.
digits

(integer) number of decimal places to round the answer to. Default is 3.
object

lm_beta object.

Value

Object of class lm_beta which is a list with 2 named fields:

  • b named numeric vector with regression coefficients (not standardized);

  • beta named numeric vector with standardized coefficients from lm() model.

Details

This function is inspired by function QuantPsyc::lm.beta() written by Thomas D. Fletcher.
coef_standardized() provides standardized coefficients even when interaction members are present. This is achieved by computing whole model matrix (with all right-hand side members of formula used in call of lm()) and calculating standard deviations of each regressor (including interaction members) based on these columns.
coef_standardized() does not fail if intercept is not present.

The remaining calculations are the same as in QuantPsyc::lm.beta().

See also

Examples


data(USJudgeRatings)
us <- USJudgeRatings
names(us)
#>  [1] "CONT" "INTG" "DMNR" "DILG" "CFMG" "DECI" "PREP" "FAMI" "ORAL" "WRIT"
#> [11] "PHYS" "RTEN"

lm1 <- lm(CONT ~ INTG + DMNR + log(DILG), data = us)
coef_standardized(lm1)
#> Standardized Regression Coefficients:
#>      INTG      DMNR log(DILG) 
#>    -0.293    -0.316     0.533 

lm2 <- lm(CONT ~ INTG + DMNR * DILG, data = us)
coef_standardized(lm2)
#> Standardized Regression Coefficients:
#>      INTG      DMNR      DILG DMNR:DILG 
#>    -0.266    -0.626     0.301     0.495 

summary(coef_standardized(lm2))
#> Summary of Standardized Regression Coefficients:
#>     regressor       coeff standardized_coeff influence_rank
#> 1 (Intercept)  9.57813871                 NA             NA
#> 2        INTG -0.32520383             -0.266              4
#> 3        DMNR -0.51532509             -0.626              1
#> 4        DILG  0.31487864              0.301              3
#> 5   DMNR:DILG  0.03270199              0.495              2


# Do not include intercept
lm3 <- lm(CONT ~ 0 + INTG, data = us)
coef_standardized(lm3)
#> Standardized Regression Coefficients:
#>  INTG 
#> 0.751 
summary(coef_standardized(lm3))
#> Summary of Standardized Regression Coefficients:
#>     regressor     coeff standardized_coeff influence_rank
#> 1 (Intercept) 0.0000000                 NA             NA
#> 2        INTG 0.9174981              0.751              1