I am currently working on a project involving a variable that represents the severity of a disease on an ordinal scale of 1 to 4, where 1 represents the least severe and 4 represents the most severe. Conceptually, this variable seems categorical in nature as it represents different ordered categories of disease severity.

The question I am grappling with revolves around the treatment of this variable. I initially planned to treat this variable as a factor (considering it categorical) to account for the distinct levels of severity, which are not necessarily evenly spaced nor do they represent a linear increase in severity.

However, my superiors have instructed me to treat this variable "as is", i.e., as a numeric variable, without converting it into a factor. This directive has left me questioning the statistical correctness and potential implications of such an approach.

My concerns are:

Is it statistically valid to treat an ordinal variable as numeric? Could this potentially lead to misleading results, given that the distance between each successive category might not represent a consistent increase in disease severity? Or is it a common practice in certain statistical analyses to treat ordinal variables as numeric? Any guidance or advice on this matter would be greatly appreciated. Particularly, I would be grateful for any references or resources that could shed light on this topic to better inform my approach and discussion with my team.



Browse other questions tagged or ask your own question.