bayesline.api.equity.TSFactorSpec

pydantic model bayesline.api.equity.TSFactorSpec

Specification for deriving a risk model with user-defined TS factors.

This spec is consumed by derive_factor_model_with_tsfactor to build a new FactorRiskModelSettings that is then run as an ordinary report.

The uploaded series is interpreted as the returns of an external portfolio or theme. Per-asset exposures are derived at report time by rolling regression of asset returns on the uploaded series; the resulting exposures feed a chained second-stage regression that produces the new factor’s returns alongside the existing model’s factors.

Show JSON schema

{
   "title": "TSFactorSpec",
   "description": "Specification for deriving a risk model with user-defined TS factors.\n\nThis spec is consumed by ``derive_factor_model_with_tsfactor`` to build a new\n``FactorRiskModelSettings`` that is then run as an ordinary report.\n\nThe uploaded series is interpreted as the returns of an external portfolio\nor theme. Per-asset exposures are derived at report time by rolling\nregression of asset returns on the uploaded series; the resulting\nexposures feed a chained second-stage regression that produces the new\nfactor's returns alongside the existing model's factors.",
   "type": "object",
   "properties": {
      "tsfactors_source": {
         "description": "Name of the uploaded time-series factors dataset to draw from.",
         "title": "Tsfactors Source",
         "type": "string"
      },
      "factor_selection": {
         "description": "Factors to pull from the uploaded dataset. Each becomes a custom factor in the derived risk model.",
         "items": {
            "$ref": "#/$defs/TSFactorSelection"
         },
         "title": "Factor Selection",
         "type": "array"
      }
   },
   "$defs": {
      "TSFactorSelection": {
         "description": "Selection of a single factor from an uploaded time-series dataset.",
         "properties": {
            "factor_name": {
               "description": "Name of the factor column within the uploaded dataset.",
               "title": "Factor Name",
               "type": "string"
            },
            "display_name": {
               "anyOf": [
                  {
                     "type": "string"
                  },
                  {
                     "type": "null"
                  }
               ],
               "default": null,
               "description": "Optional display name for reports. Falls back to factor_name when omitted.",
               "title": "Display Name"
            },
            "factor_group": {
               "default": "Thematic",
               "description": "Factor group label for grouping in reports.",
               "title": "Factor Group",
               "type": "string"
            },
            "window": {
               "default": 252,
               "description": "Rolling window length (trading days) for TS-beta exposure.",
               "minimum": 2,
               "title": "Window",
               "type": "integer"
            },
            "method": {
               "default": "huber",
               "description": "Rolling regression method for TS-beta exposure.",
               "enum": [
                  "ols",
                  "huber"
               ],
               "title": "Method",
               "type": "string"
            }
         },
         "required": [
            "factor_name"
         ],
         "title": "TSFactorSelection",
         "type": "object"
      }
   },
   "required": [
      "tsfactors_source",
      "factor_selection"
   ]
}

Config:

• frozen: bool = True

Fields:

• factor_selection (list[bayesline.api._src.equity.report.scenarios.TSFactorSelection])

• tsfactors_source (str)

Validators:

• _validate » all fields

field tsfactors_source: str [Required]

Name of the uploaded time-series factors dataset to draw from.

Validated by:

• _validate

field factor_selection: list[TSFactorSelection] [Required]

Factors to pull from the uploaded dataset. Each becomes a custom factor in the derived risk model.

Validated by:

• _validate