I'm looking at a data set where one of the variables is an ordinal scale with six possible levels. However, the specific data set ended up only having members of four of those levels. We have representatives of 1, of 2, of 5, and of 6, but no 3 or 4. No, I can't go back to the clinicians and get data to fill things in. I got what I was handed.
How do I account for the fact that there is a very real gap between the 1 and 2 samples and the 5 and 6 samples? In the real world, using this scale, we can find people who score 3 and 4, but they weren't included in the data the clinicians collected.