I see, I think I understand your question better now. The reason the example Tablet Production data set won't allow a polynomial fit is because the Screen Size data has a modeling type of Nominal. If you change that to Continuous, then a polynomial fit is possible. See the lower left portion of the image below -> change the degree of the fit to quadratic, for example, to get a quadratic fit instead of linear.
As long as your data type is numeric, and it makes sense to set the modeling type to Continuous, then a polynomial fit is possible.
(Note that you will want to use the Line of Fit graph for this, not the Smoother. These two options have similar looking icons but work very differently, especially for continuous vs nominal data.)
