dmp_multi_horizon – Diebold-Mariano-Pesaran joint multi-horizon test.#

Back to equal_predictive_test axis | Back to L6 | Browse all options

Operational op under axis equal_predictive_test, sub-layer L6_A_equal_predictive, layer l6. Standalone callable: mf.functions.dmp_test.

Function signature#

mf.functions.dmp_test(
    loss_differentials: list[np.ndarray] or np.ndarray,
) -> DMPTestResult

Parameters#

name

type

default

constraint

description

loss_differentials

list[np.ndarray] or np.ndarray

—

—

Per-period loss differentials, one array per horizon or pre-stacked.

Returns#

DMPTestResult — frozen dataclass with fit results.

Attribute

Type

Description

stat

float or None

DMP test statistic

pvalue

float or None

Two-sided p-value

decision

bool

Reject H0 at 5%

n_obs_stacked

int

Stacked observations

Behavior#

HAC-adjusted stacked DM test that evaluates equality of predictive ability across all forecast horizons simultaneously. v0.3 implementation following Pesaran-Timmermann.

When to use

Joint significance across multiple horizons (avoids per-horizon p-value adjustment).

In recipe context#

Set params.equal_predictive_test = "dmp_multi_horizon" in the relevant layer to activate this op within a recipe:

# Layer L6 recipe fragment
params:
  equal_predictive_test: dmp_multi_horizon

References#

  • macroforecast design Part 3, L6: ‘tests must report (statistic, p-value, kernel, lag) and respect HAC dependence-correction.’

  • Pesaran & Timmermann (2007) ‘Selection of estimation window in the presence of breaks’, JoE 137(1): 134-161.