Fit options
Fit options
These options are configured per model fit. The defaults, domains, and descriptions below are generated from the central option schema.
| Group | Option | Default | Domain | Description | Example |
|---|---|---|---|---|---|
| fit | order |
(1, 1, 0) |
Three non-negative integers. | ARIMA order (p, q, d) (AR, MA, integer differencing). |
(2, 1, 2) |
| fit | garch_order |
(1, 1) |
Two non-negative integers. | GARCH order (k, m). |
(2, 2) |
| fit | fractional_d |
None |
None, “estimate”, or a finite real number strictly between -0.5 and 0.5. | Conditional fractional differencing of regression residuals; estimate one common d or supply a fixed value. Requires regular observations and zero integer differencing. | estimate |
| fit | exact_ll |
None |
None or boolean. | Gaussian covariance correction. None enables it for supported ARIMA models with nonzero order and no GARCH, including panels and effects; static and unsupported models remain conditional. True forces it on; False forces conditional likelihood. Exact stationary likelihood without effects or variance regressors; finite-history approximation otherwise. | True |
| fit | vol |
'GARCH' |
'GARCH', 'EGARCH' |
Volatility model. | EGARCH |
| fit | effects |
Effects() |
An Effects instance. | Fixed or random effects in the mean and variance models. | | | fit | `optimizer` | `OptimizerOptions()` | An OptimizerOptions instance. | Numerical optimizer settings. | |
| fit | constraints |
None |
None, a string, or a dictionary. | Restrictions on coefficient estimates. | | | fit | `add_intercept` | `True` | True or False. | Add an intercept when one is not present in the formula. | |
| fit | subtract_means |
False |
True or False. | Subtract variable means before estimation. | | | fit | `tobit_limits` | `(None, None)` | Two finite real numbers or None, with lower <= upper. | Lower and upper limits for a Tobit model. | `(0, None)` | | fit | `suppress_output` | `True` | True or False. | Suppress fit status output. | |
| fit | likelihood |
None |
None or a likelihood class. | Custom likelihood model class. | | | fit | `h_function` | `None` | None or a supported h-function. | Custom heteroskedasticity function. | |
| effects | group |
'none' |
'none', 'fixed', 'random' |
Group effect specification. | | | effects | `time` | `'none'` | `'none'`, `'fixed'`, `'random'` | Time effect specification. | |
| effects | variance |
'none' |
'none', 'fixed', 'random' |
Variance effect specification. | | | optimizer | `tolerance` | `0.0001` | Positive number. | Tolerance used by the numerical optimizer. | `1e-5` | | optimizer | `max_iterations` | `150` | Positive integer. | Maximum number of optimization iterations. | `300` | | optimizer | `accuracy` | `0` | Non-negative integer. | Optimization accuracy level. | `1` | | optimizer | `initial_arima_garch_params` | `0.1` | Non-negative number. | Initial size of ARIMA-GARCH parameters. | |
| optimizer | arma_constraint |
3 |
Positive number. | Maximum absolute value of ARMA coefficients. | | | optimizer | `arma_round` | `14` | Positive integer. | Number of significant digits used in ARMA matrices. | |
| optimizer | garch_min |
1e-12 |
Positive number. | Minimum absolute value of GARCH coefficients. | | | optimizer | `garch_assist` | `0` | Non-negative number. | Weight assigned to the assisting GARCH variance. | |
| optimizer | garch_ridge_wald |
0.2 |
Finite non-negative real number (excluding booleans). | Absolute squared-Wald-ratio threshold (strict) for attempting to fix the highest free gamma lag at zero. Negative variance uses a heuristic; zero variance is skipped. Accepted reductions get one release retry within the iteration budget; 0 disables ridge reduction. | 1.0 |
| optimizer | multicoll_threshold_report |
30 |
Positive number. | Condition-index threshold for reporting multicollinearity and gating automatic ARMA/GARCH reductions together with their Wald criteria. | | | optimizer | `min_group_df` | `1` | Positive integer. | Minimum observations allowed in each group. | |
| optimizer | robust_cov_lags |
(100, 30) |
Two integers greater than one. | Lags used for robust covariance calculations. | (100, 30) |
| optimizer | variance_re_norm |
5e-06 |
Positive number. | Normalization point for variance random-effects calculations. | | | optimizer | `kurtosis_adj` | `0` | Non-negative number. | Kurtosis adjustment. | |