Say I am producing a wound coil which is ultimately going to be used in a colpitt's oscillator circuit. I measure the Ls and Rs of the coil, and then install it in a ferrite and then test it in the circuit, where I determine if it works or not, based on the resulting frequency and amplitude of the oscillator output. What is the right statistics tool to help set the production tolerances for said coils? I was thinking that I can use the measures Ls and Rs and use them as the features, and the circuit performance to label the data good or bad, and use LDA to see if the good vs, bad coils are separable, and then set the tolerance based on the projected good coil's distribution mean and SD ...is this a good approach? Is there a better way?