Hello everyone,
I want to use gender as a covariate in an ancova. I read that I have to dummy code it beforehand (male, female, other). When doing that, I would get three variables in spss, right? Do I include all of them as a covariate in the analysis or just chose one? I would be very grateful if someone could explain this to me.
Thank you,
Yara
Hello Yara Stephan. The conventional way to estimate an ANCOVA model with SPSS is to use the UNIANOVA command. Click Analyze > General Linear Model > Univariate if you want to use the GUI to generate the command syntax. In the dialog, put the Group variable (e.g., gender) in the Fixed Factor(s) box, and the quantitative explanatory variable in the Covariate(s) box. The basic syntax looks like this:
UNIANOVA y BY group WITH x /EMMEANS=TABLES(group) WITH(x=MEAN).
Replace group with your gender variable and x with the quantitative explanatory variable in your model (i.e., the covariate).
Notice that BY is followed by a list of one or more categorical variables (those that were in the Fixed Factors box) and WITH is followed by a list of one or more quantitative variables (the ones in the Covariates box in the GUI dialog).
HTH.
Dear Yara Stephan
ANCOVA, which is part of a linear family model, assesses whether the means of the variable of interest are equal between the groups of a categorical variable coded for each of its modalities (nominal or categorical).
This means that all the coded modalities 1 (male) ,2 (female) ,3 (other) should be used in the model to test for significant difference or similarity of the mean between groups 1, 2 and 3.
It is also important to rigorously check the validity of your model by testing it with statistical tests and by eliminating explanatory or predictive variables that are not of great interest as they may interfere with the analysis.
The link below is to a more detailed case study in SPSS. It is recommended to study it also with other software such as EViews, R , SAS, or STATA.
https://www.scalestatistics.com/ancova.html
Warm Regards,
Conversion of your grouping variable is only necessary if you are going to use the form of regression that is equivalent to ANCOVA. In that case, you would use 2 dummy variables to represent a 3-category variable.
- Badges
- Science topic
- Similar topics
- Physical Sciences
- Materials Science
- Plating