To understand my rationale, I've used the equation of enterprise value
http://en.wikipedia.org/wiki/Enterprise_value together with some simplifying assumptions there. (Bitcoin ecosystem is an enterprise, albeit a very unusual one).
If you construct a notional enterprise such that the equation is applicable, that's cool, but I haven't seen any such construction. It has to account for reserve demand and allow one to model the dynamics of the equation inputs in order to be remotely useful.
I hold that bitcoin most usefully modeled as a currency, that the most useful and germane equation is PQ=MV.
Various sentiment and expectation models, agent aggregations factored by constituency, may be fit, and if it is done adequately, this can offer some improvement in predictive power relative to a coin flip, but there are inherent limits in such a model, and constructing one with rigorous error bars is a costly task - as is narrowing those error bars by judicious data collection.
An influence model might incorporate these as well as growth estimates on the basis of historical comparables. SMTP is often held out as a comparable, for example.