This function identifies aliased (linearly dependent) variables in a linear model by fitting a linear model, and then using the stats::alias function to detect aliased variables.
Value
Returns a character vector of aliased variable names if any are
found; otherwise, returns NULL
invisibly. If verbose
is TRUE
, the
function will also print a message to the console.
Examples
load_packages(car)
x1 <- rnorm(100)
x2 <- 2 * x1
x3 <- rnorm(100)
y <- rnorm(100)
model <- lm(y ~ x1 + x2 + x3)
summary(model)
#>
#> Call:
#> lm(formula = y ~ x1 + x2 + x3)
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -3.1728 -0.6343 -0.1377 0.8417 2.3770
#>
#> Coefficients: (1 not defined because of singularities)
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 0.12272 0.10481 1.171 0.245
#> x1 0.17776 0.11061 1.607 0.111
#> x2 NA NA NA NA
#> x3 0.03323 0.10549 0.315 0.753
#>
#> Residual standard error: 1.029 on 97 degrees of freedom
#> Multiple R-squared: 0.02599, Adjusted R-squared: 0.00591
#> F-statistic: 1.294 on 2 and 97 DF, p-value: 0.2788
#>
# there are aliased coefficients in the model
try(car::vif(model))
#> Error in vif.default(model) : there are aliased coefficients in the model
# The function identifies the aliased variables
detect_alias(data = cbind.data.frame(x1, x2, x3))
#> [1] "x2"
detect_alias(data = cbind.data.frame(x1, x2, x3), verbose = TRUE)
#> aliased variables: x2
#> [1] "x2"
# excluding x2 and refit the model
model <- lm(y ~ x1 + x3)
summary(model)
#>
#> Call:
#> lm(formula = y ~ x1 + x3)
#>
#> Residuals:
#> Min 1Q Median 3Q Max
#> -3.1728 -0.6343 -0.1377 0.8417 2.3770
#>
#> Coefficients:
#> Estimate Std. Error t value Pr(>|t|)
#> (Intercept) 0.12272 0.10481 1.171 0.245
#> x1 0.17776 0.11061 1.607 0.111
#> x3 0.03323 0.10549 0.315 0.753
#>
#> Residual standard error: 1.029 on 97 degrees of freedom
#> Multiple R-squared: 0.02599, Adjusted R-squared: 0.00591
#> F-statistic: 1.294 on 2 and 97 DF, p-value: 0.2788
#>
try(car::vif(model))
#> x1 x3
#> 1.022448 1.022448