After correction for scattered light and sky contribution (which can be done more or less without reference to other observations), the data consist of a raw measure of the (vector) polarization of the object itself, added to an instrumental zeropoint (vector) and in a Stokes parameter system with indeterminate scale (degree of polarization) and orientation (polarization angle). These quantities are all potentially wavelength-dependent and must be calibrated by observations of `standards' of one kind or another with the instrument set up exactly as for your programme.
Which of the above properties is the most important in practice will be determined by the ratio of object polarization to instrumental zeropoint. Fig. 7a shows the general case. Note that while, for sufficient photons, and are normally distributed, P and 2 are not. In averaging equivalent observations, therefore, we should average , and compute `average' P and 2 from them. For cases such as in Fig. 7b, exact knowledge of the zeropoint is most important, for Fig. 7c scale and orientation are much more significant uncertainties. Your scientific aims can be translated into required accuracy for each of these quantities. Depending on the required accuracies, calibration may be simple and straightforward or exceedingly complex and subtle. General guide lines are given for each quantity in turn; it is up to you to devise a satisfactory procedure for your particular problem.
Having elected to use spectropolarimetry as your preferred scientific method, it is likely that you will be able to do at least part of your calibration in a spectrally differential way. When using `standard stars', be aware of the difference between classical polarimetry and what you are attempting. Certified standards for broadband work may not be suitable now; you need to devise your own consistency checks. You should also realise that, at levels below 0.1% and 1 degree, many stars have variable polarization (i.e. broadband, it may be worse spectrally). The dilemma is that bright polarized stars are generally distant supergiants with unstable atmospheres, but at the highest accuracy every star is in fact suspect (as are terrestrial sources; more so than stars, in fact).