sae.all_subsets_fh

All-possible-subsets model selection for Fay-Herriot area-level models. Enumerates every subset (up to max_predictors) of the candidate features, fits each with Fisher-scoring EBLUP, keeps the k_best per predictor count by AIC, and selects the minimum-AIC model whose coefficients are all significant at alpha. Non-converging or singular fits are skipped, not fatal. The selected model is written fitted, so its model output feeds sae.fay_herriot_mse directly.

Kind: standard
Network: no

Inputs🔗

Name	Type	Description
`data`	file (Parquet or CSV)	One row per area, with the target, candidate predictor and sampling-variance columns.
`target`	string	The direct-estimate column.
`features`	string	Column selector for the candidate pool, e.g. `B15,B20,B25,B30,B35,MB15,MB20,MB25,MB30,MB35` or `B,MB,-B5,-MB5`.
`vardir`	string	Name of the column holding each area's sampling variance.
`intercept`	boolean	Include an intercept. Default `false`.
`id_columns`	string	Column selector for identifier columns, never used as predictors. Default empty.
`categorical`	string	Column selector for numeric-coded predictors to treat as categorical. Default empty.
`na_action`	string	`fail` (default) or `drop`: what to do with an area that has a missing value in a model column.
`method`	string	Variance estimation method - one of `REML` (default), `ML`, or `FH`.
`k_best`	number	Keep the k best models (by AIC) per predictor count. Default `5`.
`max_predictors`	number	Largest subset size to consider, clamped to the number of candidates. Default `5`.
`alpha`	number	Significance threshold: a model qualifies only if every coefficient has p < alpha. Default `0.01`.
`max_iter`	number	Maximum Fisher-scoring iterations per fit; non-converging fits are retried once at twice this. Default `500`.

Outputs🔗

Name	Type	Description
`ranking`	file (CSV)	The `k_best` models per predictor count, sorted by AIC: `p`, `rank`, `predictors`, `aic`, `all_signif`, `max_pvalue`.
`model`	file (SpadeModel)	The selected model, fitted (family `fay_herriot`).
`fit`	json	The selected model's fit statistics, and `selection`: `predictors`, `aic`, `alpha`, `k_best`, `max_predictors`, `candidates`.

Example🔗

- id: "@selection"
  name: sae.all_subsets_fh
  inputs:
    - "@long_data"
  args:
    target: "estimate"
    features: "B15,B20,B25,B30,B35,MB15,MB20,MB25,MB30,MB35"
    vardir: "var_of_estimate"
    max_predictors: 3
    alpha: 0.05