When testing the accuracy of an interpolator, select the best method$($s$),$ that give the most robust estimate of error without reducing the power of your estimations too much

Question $37$ options:

A "leave one out" cross validation

Calculating an RMSE

Collecting a new, independent set of test points

Splitting your sample data into a fitting set, with about $75\%$ of the samples, and testing with a reserve set $($about $25\%)$ held back from the fits, comparing the predicted to observed values for the withheld points