Whether one independent variable in multilevel regression can be the other two independent variables, such as type, token, and type/token ratio
Dear Jiayao Ye
be careful!
It is not advisable to mix variables of different types in multiple regression analysis models
All independent variables should be exclusively quantitative to ensure the quality of the model and the validity of the method
If some variables are ordinal, they can be entered into the model,
except if they can be interpreted in ascending or descending order.
As for qualitative variables that are more than binary, this is not desirable, and experimental design tests can be used to test them, such as Ancona, Mancova, and Manova
If the independent variable is binary, Binary logistic regression can be used
It can be entered into the multiple model as a dummy variable
that takes the value of 0 or 1
Jiayao Ye
You can see the following books for more details
1. Using Multivariate Statistic by Tabachnic
2. Hair - Multivariate Analysis 7e
3. Applied Multivariate Techniques - Subhash Sharma
You can include two of those independent variables, but not all three. You'll create multicollinearity.
This is true in any regression, not just multilevel.
In multilevel regression, it's crucial to avoid perfect multicollinearity where one independent variable is a perfect linear combination of others. Using type, token, and type/token ratio as independent variables is generally acceptable if they aren't perfectly correlated. However, careful consideration is needed to prevent multicollinearity issues, as high correlations can destabilize coefficient estimates. Assess variance inflation factors (VIF) to gauge multicollinearity. If VIF values are low, including type, token, and type/token ratio as independent variables in a multilevel regression model can be viable, provided their contributions align with the research question and data characteristics.
- Badges
- Science topic