"compliance with the assumption of homoscedasticity" is also ok, but it's more common to refer to simply "testing the assumption" rather than "testing compliance with the assumption."

I also propose changing "transversal variables" to something like subject-level variables. Without having more of the source text it's hard to say what the best translation for variables transversales would be, but the text generally is referring to repeated measures variables (medidas repetidas, e.g. if someone's blood pressure was measured daily for a month) and variables that are not measured repeatedly (variables transversales, e.g. someone's date of birth or blood type, things that don't change over time). These non-repeated measures are often referred to as subject-level variables, but depending on the source text there may be a better option (e.g. person-level variables, if the study is about people).