Profiler with non-negative numeric response data -- how to set zero boundary?
Aug 6, 2020 10:44 AM
| Last Modified: Aug 6, 2020 10:45 AM(174 views)
In a test, three components are weighed before and after a treatment. The components all gain weight, due to the nature of the test. It is not possible for a component to lose weight. Various factors of interest may have caused the weight gain. In a typical model, one can choose to see the Profiler. Here, without constraint, the Profiler can predict a negative response for a given factor level.
Is there a way to constrain the Profiler to avoid suggesting a negative response for a factor level? Please offer guidance!
The profiler is only you the predicted outcomes from your model. What you need to do is create a model which doesn't predict meaningless outcomes. Typically you achieve this by applying a transformation to your response variable.