Can we double check the results of an already obtained clear and significant multiple logistic regression models estimating cancer mortality within a specific period of time (for example, 2005-2010), by creating a new validation set for a more recent period of time (for example, 2011-2014)? Will that add any value to an already established and sound statistical analysis based on the prior period?
Temporal validation will help reinforce the robustness of the model . It will also help identify any deviations/changes in the risk factors.
Traditional epistemological approaches (Karl Popper) suggest that we can never be sure that our hypotheses are true. Instead, a theory (or hypothesis) can be considered valid as long as it has not (yet) been falsfied. Moreover, the creditbilty of the theory or hypothesis will increase with the number of failed attempts to prove it wrong. So, my answer to your questions is a clear yes. Testing your model with new data will strengthen your hypothesis or theory, no matter how perfect the statistical modeling was. Moreover, spliting data into training and test data sets has become a standard approach in many statistical domains, e.g. forcasting as well as machine learning (predictive modeling). Model performance in latter approaches is usually assessed by comparing RMSE, R2 or similar measures and you may want to do the same with your two data sets.
- Badges
- Science topic
- Similar topics
- Medicine
- Physiology
- Biosignals
- Biomedical Signal Processing
- Statistical Analysis