Hello everybody,
I'm doing a multilevel analysis by using the statisical software HLM 7. At level 1, I have two nominal variables with 3 categories each. So far, I could only find information how to include dichotomous or metric variables as predictors. I tried to recode one variable in 3 dummy variables, but that didn't work due to multicollinearity.Does anyone have an idea how to include variables as predictors with 3 categories?
Thanks in advance!
3 categories are represented as 2 dummies (and a constant). You are the modelling differentials in the outcome from the estimate of the base or reference Category that has been omitted , thereby avoiding exact collinearity. Exactly the same as in standard multipl regression.
You could use the "effect coding" approach.
It is described here
https://www.researchgate.net/publication/256349393
Dear Kelvyn and Gaëtan, thank you very much for you replies! The article seems very helpful! Best wishes, Josefine
This might also help
https://www.researchgate.net/post/Why_does_SPSS_exclude_certain_variables_from_a_regression
and this
https://stats.idre.ucla.edu/spss/faq/coding-systems-for-categorical-variables-in-regression-analysis/
which sets out in reasonable detail a number of different schemes.
K
- Badges
- Science topic
- Similar topics
- Agriculture