Pspice allows R and G to be functions of a laplace variable [frequency]. In
theory, skin and dielectric loss can be modelled if the correct function of R
and G are chosen (in my experience, G must be set much higher than actual to get
correlation, but this is another question !). Alternately, a frequency dependent
two port representation can be created using Pspice's analog behavioural
modelling suite. In this case, the two port might be described in terms of
scattering (this is outlined in Pspice Apps notes) or ABCD matrices and
implemented with analog behavioural modelling components.