Which of the following is NOT an advantage of regression trees over linear regression?

The assertion that regression trees have a higher level of predictive accuracy than linear regression is not universally true and can depend on the context and the specific dataset being analyzed. While regression trees can capture nonlinear relationships and interactions within the data, leading to improved performance in certain scenarios, it is not guaranteed that they will always outperform linear regression in terms of predictive accuracy.

In contrast, the other options highlight key advantages of regression trees. For instance, regression trees are often easier to interpret since they split the data into subsets based on predictor values, allowing for straightforward examination of decision paths. They can be visually presented in a tree format, making the model structure and decisions clear. Additionally, regression trees handle qualitative predictors inherently by treating categorical variables as split points, eliminating the need for preprocessing steps like creating dummy variables, as is required in linear regression.

Thus, the statement regarding predictive accuracy is the outlier, as the performance of regression trees versus linear regression can vary based on the characteristics of the data involved.