I think you might be better off code-modeling the various
functions and artifacts of interest, in a more straightforward
and expressive way. Why would you want to do it using
polynomial math, when you can talk sample time and slope
error, sample pedestal, through-delay and rise-fall? I bet
there's an elegant one-liner that's like a koan or something,
if you only wanted the event value and not the details.
There's probably verilog-a code out there already that you
could dig up, rip up and turn into MATLAB or whatever.