# Can I treat ordinal values (with an underlying order) as continuous when extracting residuals and in exploratory factor analysis?

I am looking for clarification on whether it is acceptable to treat ordinal values with an underlying order as continuous when extracting residuals and in exploratory factor analysis (EFA).

1. Does taking the residuals from a linear regression make sense if the independent variable is ordinal (ie: 1,2,3,4,5,6,etc) but has an underlying order? Or, do I need to use a regression model specific to ordinal values?
2. Is it possible to perform an EFA on a dataset that has both ordinal and continuous values in it by treating all values as continuous?
3. If the latter is not advisable, is there a method that would allow me to complete my analysis (taking residuals, performing EFA) on a mixed dataset of continuous + ordinal variables?

Thanks in advance- any help much appreciated!

• You went astray the moment you categorized your trait variable in an effort to cope with its censoring at both ends: you both complicated the analysis and lost information, which will make the results worse than you might otherwise achieve. There are better ways. Would you like to change your question to ask how to accomplish your original aim, or do you still want to pursue the path you have taken? – whuber Oct 3 '17 at 17:39
• @whuber, thanks for your response. I am still curious to know how to deal with datasets with heterogeneous variable types, so I have made my original question more general. Yes, I am very interested in finding out better ways to accomplish my aim! Did some research after reading your comment; interval censoring regression appears appropriate for my dataset. Is this the method you were referring to? – Yedigei Oct 4 '17 at 15:47
• Just explore posts having tag ordinal data. – ttnphns Oct 4 '17 at 17:25
• Yes, "interval censoring" is a good search term. – whuber Oct 4 '17 at 17:44