I am currently working with a data set where I would like to answer the research question using a linear mixed models approach. I would like to introduce 2 categorical covariates into the model, but do not know how to code this.
I have continuous dependent variables, I'm using the subject as the random factor, and I have a dichotomous (win versus loss) fixed factor.
Code this like any other independent variable. Best wishes David Booth
Hello Jesse A Stein. Presumably you are using the MIXED command. Here is the Command Syntax Reference documentation for MIXED:
https://www.ibm.com/docs/en/spss-statistics/28.0.0?topic=reference-mixed
Notice the first line in the syntax chart:
MIXED dependent varname [BY factor list] [WITH covariate list]
In SPSS lingo, factor variable = categorical explanatory variable, and covariate = quantitative explanatory variable. (If you are using the GUI to generate your code, you'll see boxes for factors, or maybe fixed factors, and for covariates.)
If you scroll down a bit and click on the Variable List link, you'll find some more info as follows:
The variable list specifies the dependent variable, the factors, and the covariates in the model.
- The dependent variable must be the first specification on MIXED.
- The names of the factors, if any, must be preceded by the keyword BY.
- The names of the covariates, if any, must be preceded by the keyword WITH.
- The dependent variable and the covariates must be numeric.
- The factor variables can be of any type (numeric and string).
- Only cases with no missing values in all of the variables specified will be used.
HTH.
PS- Dichotomous explanatory variables can be treated as either fixed factors or as covariates. If you treat them as covariates, 0/1 (indicator) coding is convenient. But for categorical variables with 3 or more levels, it is easier to treat them as fixed factors.
David, when you said, "Code this like any other independent variable", I assume you were talking about the case of dichotomous variables. Right?
As you well know, I am not a useR (which is why I try not to answer questions about how to do X using R). But in the case of categorical explanatory variables with 3+ levels, I should think that even useRs do not code them exactly the same as quantitative variables. Surely they would use the factor function or one of the other approaches shown on this UCLA page, right?
- https://stats.oarc.ucla.edu/r/modules/coding-for-categorical-variables-in-regression-models/
Thank you for clarifying. And if I completely misunderstood what you meant, I apologize.
- Badges
- Science topic
- Similar topics
- Philosophy
- Esthetics
- Interpretation