Hello beautiful scientists:
How are you? So I have the above mentioned question. Papers are not clear in the field of adsorption whether to include this in the manuscript or not.
So, I did a regression, did an ANOVA, and some terms are insignificant, so those were removed. And now the interactive effects are not included in the regression equation. This I understand to be due to predicting a response, and also to see if the response matches the experimental values, and how far apart they are (via residuals).
So, should I still do contour graphs or 3D plots with nonsignificant interactive effects for my manuscript?
Thank you
In my opinion there is no need for that, just go with plotting of the final, refined model, it is the one that is providing better approximation to the actual response surface over your experimental space, than the overdefined model with all regressors included.
Absence of evidence is no evidence of absence.
It is not reasonable to remove a term from a model just because it is not statistically significant. So if you had a reason to include a term, then it should be part of the model, no matter if the data was sufficient to get a smallp-value for the coefficient.
I don't think that a surface plot is neccesary to check the model - residual diagnostic plots are sufficient for that purpose. But disregarding terms just because the p-values of their coefs are larger then some threshold is (usually) not ok.
- Badges
- Science topic