If you have a model of your system, know the poles and zeros, either method is fine. If you are more comfortable with the complex plane, use the root locus method. If you prefer working with bode plots then you may want to use the Nyquist stability criteria method.
If you do not have a model of your system, but have measured frequency response data of the system then you can still use the Nyquist stability criteria method without having a model.
Hope this helps.