§ m methods

Twenty estimators,
each one seriously.

Not an estimator menu over a generic regression call. Each method has its own assumptions, diagnostics, and defaults — surfaced in the UI, printed in the output, citable in your paper.

Panel · dynamic · Pesaran, Shin & Smith · 1999

Pooled Mean Group ARDL

Long-run homogeneity across panels with short-run heterogeneity. Appropriate when regressors are I(1) and cointegrated with the dependent variable.

Δyᵢₜ = φᵢ (yᵢ,ₜ₋₁ − θ′xᵢₜ) + Σⱼ γᵢⱼ Δxᵢ,ₜ₋ⱼ + αᵢ + εᵢₜ

Eq. — Canonical specification. Configure the ARDL order, cluster structure, and SE type in the inspector.

Specification sheet

Sample: Medium-to-long panels · N ≥ 20, T ≥ 20
Assumes: Cointegration; cross-sectional independence (or use CS-ARDL)
Output: Long-run θ (pooled); φᵢ, γᵢⱼ (panel-specific); ECT speed
Diagnostics: Hausman (PMG vs MG), Pesaran CD, Kao/Pedroni cointegration
Citation: Pesaran, Shin & Smith (1999), JASA 94(446), 621–634

Read the PMG-ARDL chapter →Download SOP PDF

§ f families

Seven families.
Twenty methods.

Linear

OLSGLSWLSQuantile

Heteroskedasticity-robust and cluster-robust SEs native.

Instrumental

2SLSIV-GMMLIML

Weak-IV diagnostics (Stock-Yogo, Montiel-Pflueger).

Panel · static

FEREBEFD

Hausman, Breusch-Pagan LM, and Mundlak decomposition built in.

Panel · dynamic

PMG-ARDLMGDiff-GMMSys-GMM

Arellano-Bond AR(1)/AR(2) tests, Hansen-J on the instrument set.

Limited dependent

ProbitLogitTobit

Marginal effects with delta-method or bootstrapped SE.

Time series

ARIMAVARVECMGARCH

Cointegration testing; IRFs with analytic or bootstrap bands.

Diagnostics · PMG-ARDL

auto-run after every fit

Test	Statistic	p-value	Verdict
Hausman PMG vs MG	3.41	0.182	PMG preferred
Pesaran CD	1.21	0.227	No CD
Kao cointegration	−4.38	0.000	Cointegrated
Breusch-Pagan LM	328.4	0.000	Heteroskedastic
LLC unit root	−2.11	0.017	Stationary

Fig. 4 — Diagnostics pane for a PMG-ARDL run. Each verdict links to the canonical reference.

§ d diagnostics

Tests, not vibes

Every fit comes with its own referee.

QuantFit auto-runs the diagnostic battery appropriate to the estimator — not a generic R²/AIC dump. Each verdict cites the canonical paper.

When an assumption fails, the panel tells you which and offers the standard remedy (robust SE, cluster, alternative estimator). Nothing silent.

Twenty estimators,each one seriously.

Pooled Mean Group ARDL

Seven families.Twenty methods.

Every fit comes with its own referee.

Twenty estimators,
each one seriously.

Seven families.
Twenty methods.