Polynomial Models and Multicollinearity

Organized by textbook: learncheme.com/ The spreadsheet can be found at learncheme.com...
Made by faculty at the University of Colorado Boulder, Department of Chemical & Biological Engineering.

[@privacyvalued4134]

1:05 It would be helpful to know what interaction terms are, when to use them, and when not to use them.

[@AbhinavSingh-oq7dk]

shouldn't each feature be standardized/normalized using mean of that particular feature? You just multiplied them exponentially. In squared feature, there are no negative values, ie values lesser than the mean.

[@privacyvalued4134]

The STANDARDIZE() function in Excel does the same thing for a single term. I thought it was an odd function for Excel to have such a function since it incurs function call overhead just to do a single subtraction and division but makes sense now in the context of dealing with multicollinearity. This video seems to say that when someone plans on doing anything other than straight multiple linear regression, it is important to STANDARDIZE() all Xn inputs beforehand to avoid multicollinearity.