Please, I am trying to do a project work using logistic regression. I need to test for multicollinearity for the independent variables using R. I installed mctest and multiColl package already. I've imported my datasets and specified my x and y. but when i want to run the test, It brings an error in match.arg(type)= arg must be NULL or a character vector. Please solutions will be much appreciated. Thank you
You can use the "glm" package in R to run logit regression, see this tutorial here:
https://stats.idre.ucla.edu/r/dae/logit-regression/
Then you can check the multicollinearity of your model by using the "check_collinearity" function in the "Performance" package:
https://rdrr.io/cran/performance/man/check_collinearity.html
For example:
model1 = glm(dv ~ iv1 + iv2, data = df, family = "binomial")
check_collinearity(model1)
Hello Deborah Ayandipe,
As far as I know you can also test for multicollinearity by using the variance inflation factor (vif). The vif assesses how much the variance of an estimated regression coefficient increases if your predictors are correlated.
A vif between 5 and 10 indicates a high correlation. If the vif is higher than 10, you can assume that the regression coefficients are poorly estimated due to multicollinearity.
To solve this problem, you can exclude one or more predictor(s) with high multicolinearity.
Another possibility might be to group highly correlated variables by using principal component analysis for instance.
Check out this function: https://www.rdocumentation.org/packages/regclass/versions/1.6/topics/VIF
I hope this helps
Best regards
Paul
Since a logistic regression model is a linear model ( linear in the regression coefficients) all the usual stuff works. Best, David Booth
- Badges
- Science topic
- Similar topics
- Vectorization