wallyxie/CON_inference: v1.1.0 "release" of Stan code producing deterministic CON inference results in Figure 5 of "A framework for variational inference and data assimilation of soil biogeochemical models using state space approximations and normalizing flows."

Changed observation noise structure. NUTS likelihood estimation and posterior prediction computation now follows y[i, ] ~ N(x_hat[i, ], 0.1 * x_hat[i,]) to y[i,] ~ N(x_hat[i,], 0.1 * obs_sd[i]), where obs_sd[i] = mean(y[i,]). This exactly matches the observation noise structure in the counterpart flow VI scheme, whereas previously, there was a modest difference in observation noise structure.

Additionally, changed R code calling Stan model to save theta and x posteriors consolidated across all NUTS chains together into a CSV file.