You can also quantify the error introduced by physical models. For example, figuring out the amplitude of the line noise on your bank of capacitors and op-amps that are making up an analog computer. Electrical engineers know (and knew) a great deal about noise in their analog systems.

Here are some cool analog computers: http://www.analogmuseum.org/english/impressions/

You might be thinking of numerical error, which rarely dominates the full uncertainty if one is doing the computation right.

You can quantify the full uncertainty in both scale/analogue models and numerical models. In my experience the full uncertainty is much more often quantified in scale and analogue models than numerical models, by the way.

Analogous to numerical error, if the systems under study differ slightly you can quantify the error introduced from the "distortion" as it is often called in the literature.