Quadratic model of RSM having some terms that are not significant can i use backward regression to eliminate those terms to make the model have better goodness of fit? Is this procedure acceptable by journals with high standard?
Well p-value selection works very poorly so I wouldn't recommend your suggestion I am sending some papers of ours that give an introduction to this topic. If you decide to pursue it I would recommend you using adaptive lasso. References and programs can be found here:
https://www.researchgate.net/post/Response_Surface_modeling_with_insignificant_p_values
https://www4.stat.ncsu.edu/~boos/var.select/lasso.adaptive.html
Examples are in our papers. Best, David Booth
Obafemi -
I'm not sure of your model, but if you have many regressors, you could be overfitting to your particular sample. If you can do some form of "cross-validation," that could be a helpful indicator.
Also, note that if you have a good combination of regressors, you may have heteroscedasticity, which naturally occurs. Basing regression weights on the coefficient of heteroscedasticity can handle that.
Best wishes - Jim
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression