One data point may not be the whole story. Perhaps you should look at the expected range of input and get several data points that are spread out within this range. A curve fit or a simple linear regression would give you a better idea for which method is better suited for your situation.

It really depends on how good your models are for the two methodologies.  In either case, you should be able to compare the fit of each model to the data you used, and some form of regression coefficient would be the benchmark for the "goodness of fit" for each model.

However, considering how far away from "perfect" your results are, comparing between the two methods seems somewhat irrelevant to me; trying to claim that the result that's 5% closer is "better" when both are more than 70% off seems to be barking up the wrong tree.

TTFN

FAQ731-376: Eng-Tips.com Forum Policies

Thankyou for your replies. Both very constructive.

I do see your point IRStuff. It is quite pointless on reporting the % better figure with that kind of accuracy in the models.