Curve Fitting and Regression Modeling vs Hypothesis Testing

4 Curve Fitting and Regression Modeling vs. Hypothesis Testing 

There has been a question about which method is more appropriate for the analysis of toxicity data. In order to make a selection it is important to understand that toxicity data is used for hazard or risk assessment. Curve-fitting and regression modeling has clear advantages. 

Hypothesis testing in the determination of NOELs and LOELs also has drawbacks largely related to the assumptions necessary for the computations. These characteristics have been listed by Stephan and Rodgers (1985) and compared to curve-fitting models for the estimation of endpoints. 

Regression after a Log Transformation Test Example Number 3 

use of typical hypothesis-testing procedures that clearly state the a value (typically 0.05) leave the (3 value unconstrained, and this skews the importance of reporting the toxic result. In other words, the typical test will be conservative on the side of saying there is no toxicity even when toxicity is present. 

---