This Python function dm_test implements the Diebold-Mariano Test (1995) to statistically test forecast accuracy equivalence for 2 sets of predictions with modification suggested by Harvey et. al (1997).

This is the codes for Diebold-Mariano test based on Python

Diebold-Mariano statistical test

Easily evaluate your forecasts with (multivariate) Diebold-Mariano and (multivariate) Giacomini-White tests of equal predictive ability and MCS.

DM_Test

Diebold-Mariano test for predictive accuracy and other useful forecasting tests

:exclamation: This is a read-only mirror of the CRAN R package repository. multDM — Multivariate Version of the Diebold-Mariano Test. Homepage: https://CRAN.R-project.org/package=multDM

An end-to-end Automated ML pipeline for empirical asset pricing & DJI forecasting. Bridges econometric rigor with modern AI using H2O AutoML. Features include advanced preprocessing (Winsorization, ADF), statistical validation via the Diebold-Mariano test, and model explainability using SHAP values. Optimized for reproducible quantitative research.

A univariate LSTM model to predict IBEX 35 index stock market returns

No description available

No description available

Dynamic Nelson-Siegel model with Kalman filter and Diebold-Mariano test.

This is a mirror of the CRAN R package repository with a paper added. multDM: Multivariate Version of the Diebold-Mariano Test. https://CRAN.R-project.org/package=multDM

MiQ-MCP: A laptop-friendly experiment for minute-scale probabilistic forecasting of NIFTY-50 returns, featuring quantile regression, Mondrian conformal prediction, and finite-sample coverage guarantees. Includes rigorous evaluation with purged CV and Diebold-Mariano tests.

- Built loss function to evaluate various potential situations. - Removed exponential trend in the data and analyzed autocorrelation and partial autocorrelation. - Performed Augmented Dickey-Fuller test on three key variables to check for stationarity. - Tested for exogenity between variables using Granger Causality Wald test. - Created Univariate and multivariate forecasts using Vector Autoregression and Autoregressive-moving-average models. - Selected the best forecasting model as the one with the minimum loss. - Used Diebold-Mariano test to determine the statistical superiority of said forecast.

Interpretable time-series forecasting on the AirPassengers dataset using ARIMA, XGBoost, LIME, and SHAP

Replication of the Diebold-Mariano forecast prediction test from their 1995 paper, evaluated on stock market volatility

No description available

No description available

Comparative machine learning analysis for housing price prediction using OLS, LASSO, and Ridge regression with forecast evaluation and Diebold–Mariano testing.

Benchmarking seven forecasting approaches on AgriPriceBD — a novel daily agricultural commodity price dataset of five commodities for Bangladesh. Covers BiLSTM, Transformer, Time2Vec ablation, Prophet and Informer failure analysis, with Diebold-Mariano significance testing.

A comparative volatility forecasting project combining econometric and deep learning models. Implements GARCH(1,1), HAR-RV, and LSTM (with walk-forward retraining) to predict market volatility using historical stock returns. Includes Diebold–Mariano tests, accuracy metrics, and visualized performance results.

Educational/portfolio project to monitor equities and forecast prices from log returns. Includes leak-free scaling, an LSTM with cheap features (trend/vol/ momentum), optional MIMO (H horizons at once), price reconstruction, comparisons to naive and SMA(5), price metrics (RMSE/MAE/MAPE, MASE), and Direction/Diebold–Mariano tests.

End-to-End Python implementation of Shin (2026)'s evaluator-locked agentic loop for transparent empirical research. Combines LLM-driven specification search with immutable evaluation harnesses, penalized regression (peLASSO), and Diebold-Mariano testing on ECB forecast data. Addresses the "garden of forking paths" crisis in AI-driven economics.

Volatility & Risk Forecasting for PRY.MI, G.MI, and REC.MI. Implements GARCH & GJR-GARCH (Gaussian/t) models in R to estimate 5% VaR. Validated via Hit Tests and Diebold-Mariano (HAC) against RiskMetrics. Results highlight the superior performance of Student’s t-models in capturing tail risk, leverage effects, and distributional asymmetries.