If you can influence data collection, question is whether you can change y. Is there a true ceiling effect? (As in you want to measure a skill most people achieve, or is it an effect of measurement). This also affects how i would model. Is the "ceiling" a real thing or not?
