I have a question around stock portfolio modeling specifically the development of the model. Is there “too much” simulating via Monte Carlo?

I’ve used Monte Carlo to simulate stock prices. I then used Monte Carlo to simulate returns based off of those stock price means and standard deviation.

Now I would like to simulate different dollar distributions to each of those stock tickers. I’m new to finance and I’m not sure if I may just be compounding errors up while I try to get to answer the question of how much money do I put into each stock identified.

I am not a financial expert, but you could apply all types of Markowitz-Portfolio or Value-at-Risk Portfolios apply to your generated data. This is however difficult to do inside in Stan, because it usually requires a quadratic solving mechanism. I don’t say its impossible, but I would first think what features of Stan is a need, otherwise I would look for a solution outside of Stan.

It sounds to me like you’re goal seeking rather than fitting a distribution. I’ve no doubt Stan could be hacked to do this but IMHO Stan really excels in fitting distributions you can write down as probabilistic models.

For instance you could simulate stock data and recover the parameters of the sim using a model fitted in Stan but that won’t tell you how to invest you money optimally across a portfolio . As Andre said you’d then need to do the actual optimisation step in the generated quantities block yourself somehow and if you did Stan could give you a distribution over your allocations at the end for example.