Mixed frequency

Back to Models and Features

Mixed-frequency models combine series sampled at different frequencies, for example monthly predictors for a quarterly target, through MIDAS and dynamic factor designs.

Pass any model string below as Arm(model=...). Extra names an optional dependency, Scaling flags whether predictors should be standardized, and Tunable counts the hyperparameters the search space exposes.

Model string	Description	Input	Extra	Scaling	Recommended preprocessing	Tunable
dfm_mixed_mariano_murasawa	Mixed-frequency dynamic factor model using Mariano-Murasawa quarterly aggregation.	panel	none	no	pass a native mixed monthly/quarterly panel from macroforecast.data.combine(…, frequency=‘native’), keep quarterly flow variables on their observed quarterly dates; the model applies Mariano-Murasawa aggregation	2
dfm_unrestricted_midas	Composite DynamicFactorMQ factors plus unrestricted MIDAS forecast head.	panel	none	no	pass a native mixed monthly/quarterly panel with column-level frequency metadata, use feature_engineering.mixed_frequency_lags directly when you need full manual control	5
midas_almon	Fixed-shape MIDAS over lag groups using midasr::nealmon-style normalized exponential Almon weights.	supervised	none	no	default	1
midas_beta	Fixed-shape MIDAS over lag groups using midasr::nbetaMT-style beta weights.	supervised	none	no	default	2
midas_step	Fixed-shape MIDAS over lag groups using normalized midasr::polystep-style step weights.	supervised	none	no	default	2
restricted_midas	midasr::midas_r-style nonlinear restricted MIDAS over explicit lag columns.	supervised	none	no	default	0
unrestricted_midas	Unrestricted MIDAS over explicit lag columns.	supervised	none	no	default	1

Reference

• Models reference page for ModelSpec, ModelFit, and fit conventions.