Skip to content

Estimation Tab

The Estimation tab fits the active model's parameters to observed data using maximum likelihood, maximum a posteriori, or Markov chain Monte Carlo.

Estimation Modes

Select one of three methods:

Method Description
MLE Maximize the Kalman-filter log-likelihood. Priors are not required.
MAP Maximize the log-posterior. Each estimated parameter requires a prior.
MCMC Sample the posterior with a random-walk Metropolis sampler. Each estimated parameter requires a prior.

MLE and MAP expose a maximum-iteration control for the optimizer. MCMC exposes the draw count, burn-in, thinning, seed, proposal scale, proposal adaptation, and the posterior point summary (mean, MAP, or last draw).

Providing Data

Enter the observable column names, then supply one column of observations per name. Values accept newline-, comma-, space-, or semicolon-separated entries. Import CSV fills the columns from a CSV whose headers match the observable names.

Configuring Parameters

The Parameters panel lists every model parameter. Select a parameter to edit it in the Estimation Details panel:

  • Estimate toggles whether the parameter is included in the fit.
  • Initial sets the starting value; Lower and Upper set optional bounds.
  • Prior (MAP and MCMC only) selects a distribution, its parameters, and an optional transform.

At least one parameter must be selected, and the estimated parameter names must be unique.

Running and Inspecting Results

Select Run Estimation to fit the model, or Estimate & Solve to also write the fitted values back to the active model and solve it.

The Latest Result section reports the method, success flag, and fitted values together with the log-likelihood, log-prior, and log-posterior. MCMC runs additionally display the acceptance rate, posterior summaries, the log-posterior trace, and per-parameter posterior distributions.

Persistence and Clearing

The selected method, run settings, data columns, and parameter configuration persist across browser refreshes. Clear resets the workspace to its initial state.

Local Persistence

Estimation workspace state is stored in the browser used to access the localhost GUI. It is not written into the model configuration.