Work

Using Conditional Linear Regression Trees to Identify Potential Construct Validity and Measurement Error Differences in Health Outcomes from a U.S. National Survey