How to calculate VIF (variance inflation factor) with sdmTMB models

[seananderson]

Question by email:

I tried to check VIF (variance inflation factor) with sdmTMB models but a few packages that I tried didn't support sdmTMB. Is there a way to check VIF?

[seananderson]

Although there's no built-in method currently, it's fairly simple to calculate yourself on the input data frame.

From the Wikipedia entry:

where $R^2_i$ is the coefficient of determination of the regression equation in step one, with $X_{i}$ on the left hand side, and all other predictor variables (all the other X variables) on the right hand side.

Here's a little function to do that and an example:

vif <- function(x) {
  v <- vapply(seq_along(x), function(i) {
    rsq <- summary(lm(x[[i]] ~ . , data = x[,-i, drop = FALSE]))$r.squared
    1 / (1 - rsq)
  }, FUN.VALUE = numeric(1L))
  setNames(v, colnames(x))
}


# subset your data frame to include just the predictors in the model:
d <- mtcars[,c("mpg", "cyl", "disp", "drat")]
vif(d)
#>      mpg      cyl     disp     drat 
#> 4.257023 6.388608 6.349561 2.147019

^{Created on 2023-04-25 with reprex v2.0.2}

[ndusit]

Dear Sean,
I found the 'performance' package that could be calculated VIF.

library(performance)
check_collinearity(M31)

Check for Multicollinearity

Low Correlation

    Term  VIF       VIF 95% CI Increased SE Tolerance Tolerance 95% CI

sDist2vill 1.03 [1.00, 1.76] 1.02 0.97 [0.57, 1.00]
sCorn50m 1.00 [1.00, 1.10e+12] 1.00 1.00 [0.00, 1.00]
sDist2Water 1.03 [1.00, 1.70] 1.02 0.97 [0.59, 1.00]

Regards,
Dusit