I'm testing biological sex (male; female) by age group (18-25; 26-34; 35-49; 50-64) interactions in the NSDUH dataset using the complex sampling version of SPSS - and without the sex by age group interaction terms, my models run and do not have a singular covariance structure. I created a dummy-coded interaction term that I consider to be a nominal variable (e.g., males 18-25; females 18-25; males 26-34; females 26-24; etc). When I add this interaction term to the models, the sex and age Wald F values are no longer computed and I get an error message that says my covariance structure is singular, the validity of results is uncertain. Would you always get this error message with interaction terms, since by nature interaction terms are a combination of two variables in your model and would be dependent on other variables?