Can the weighted model be considered the full model and the unweighted model the reduced model? Or should the structure of the full model contain a more complex term (e.g., polynomial) for an appropriate comparison in terms of lack of fit?
Yes, weighted and unweighted linear regressions can be compared to assess lack of fit in bioanalytical method validation. The weighted model often provides better fit when variance isn’t constant, which is common in such data. Comparing both helps identify the best approach for accurate quantitation. Fit quality (e.g., residuals or statistical tests) should guide model selection
Both models are valid, but different. One is not full and the other not full. It's just that weighted models do not give equal importance to all data points :)
- Badges
- Science topic
- Similar topics
- Statistics
- Survey