Skip to main content

Cookies on Macro by Mark.We use essential cookies for sign-in, locale, and security. With your permission, we also use optional cookies for preferences, for analytics, and for advertising measurement with the Meta (Facebook) Pixel. Read our cookie policy.

Macro by MarkMacro by Mark
  • Home
Loading…
Loading…
← Models
MacroConceptsPolicyModels

Smooth transition regression Model

Regression with gradual regime weights, allowing coefficients to transition smoothly as a threshold variable changes.

OverviewGraphMathReferences

Graph workspace

Transition function and regime-weighted fitted values (2x1 panel)

Graph workspace

The transition function G(st;γ,c)G(s_t; \gamma, c)G(st​;γ,c) is the defining feature. The top panel plots GGG against the transition variable to show where the economy sits on the regime spectrum. The bottom panel shows the fitted time series with color coding by GGG value, making the gradual regime shift visible over time.

Chart axis metadata
xTransition variable / Time indexstandardized units / periods · -3 to 3 (standardized) / 0 to TyTransition function / Fitted valueprobability (0 to 1) / variable units · 0 to 1 / data-dependent
Transition function across the
transition variable static chart preview showing Transition weight, Regime midpoint
Series with its regime-weighted fit static chart preview showing Simulated series, Regime-weighted fit
Regime weight over time static chart preview showing Regime weight, Regime midpoint
xTransition variable / Time indexstandardized units / periods · -3 to 3 (standardized) / 0 to TyTransition function / Fitted valueprobability (0 to 1) / variable units · 0 to 1 / data-dependent

Current results

Transition zone fraction

Explain: Transition zone fraction
T−1∑t=1T1(0.25<Gt<0.75)T^{-1} \sum_{t=1}^T \mathbb{1}(0.25 < G_t < 0.75)T−1∑t=1T​1(0.25<Gt​<0.75)

Share of observations in the transition zone. High values mean the economy spends most of its time between regimes, validating the smooth specification.

0.3100

proportion

Effective AR range

Explain: Effective AR range
∣β1(2)∣=∣(β1(1)+β1(2))−β1(1)∣|\beta_1^{(2)}| = \bigl|(\beta_1^{(1)} + \beta_1^{(2)}) - \beta_1^{(1)}\bigr|∣β1(2)​∣=​(β1(1)​+β1(2)​)−β1(1)​​

Width of the range the time-varying AR coefficient spans, from β1(1)\beta_1^{(1)}β1(1)​ when G=0G = 0G=0 to β1(1)+β1(2)\beta_1^{(1)} + \beta_1^{(2)}β1(1)​+β1(2)​ when G=1G = 1G=1: one number, the distance between the two regime coefficients. A wide range indicates strong nonlinearity; a narrow range means the linear model is adequate.

0.5000

coefficient units

RMSE improvement over linear

Explain: RMSE improvement over linear
1−RMSESTR/RMSElinear1 - \text{RMSE}_{\text{STR}} / \text{RMSE}_{\text{linear}}1−RMSESTR​/RMSElinear​

Fractional reduction in root mean squared error of the STR, evaluated at its true parameters, against the best-fitting linear AR(1) on the same sample. Values below 5 percent suggest the STR adds little; a NEGATIVE value means the least-squares linear AR fit this sample slightly better than the true nonlinear model. The linear fit is chosen to minimize this sample's error and the true model is not, so whenever the nonlinearity is small against the noise the sign can go either way: regimes that barely differ, or a gradual switch on an unlucky sample, as at the defaults and in the Gradual Phillips curve preset (about -0.4 and -0.2 percent at 300 observations).

-0.0040

signed fraction (negative = worse than linear)

Transition midpoint location

Explain: Transition midpoint location
ccc

The value of sts_tst​ where G=0.5G = 0.5G=0.5, as set on the page: this toy evaluates the STR at its true parameters, so nothing is estimated and the reported midpoint is ccc itself. The economic interpretation depends on what sts_tst​ represents (e.g., an output gap of zero, a credit spread of 200 bps).

0.0000

transition variable units

Controls

Switches gradually
Explain: How sharp the switch is

Transition speed: 3

How quickly the model moves from one regime to the other; large values make it a jump.

Controls the steepness of the S-curve in the top panel. At low values the curve is nearly flat; at high values it snaps vertically.

Both regimes visited
Explain: Where the switch happens

Location parameter: 0

The level of the transition variable at which the model sits halfway between the two regimes.

Moves the vertical dashed line and the S-curve horizontally, changing which observations fall in each regime.

Shocks fade steadily
Explain: Shock memory in regime 1

AR(1) coefficient, regime 1: 0.3

How long a shock keeps affecting the series while the model sits in regime 1.

Changes the decay rate of the fitted series during blue (regime 1) periods.

Shocks fade steadily
Explain: Shock memory in regime 2

AR(1) coefficient, regime 2: 0.8

How long a shock keeps affecting the series while the model sits in regime 2.

Changes the decay rate during red (regime 2) periods. Near-unit-root values produce sluggish dynamics.

300
Explain: How many periods simulated

How many observations the simulation generates; longer runs show more switches between regimes.

Extends the bottom-panel time series. More data makes the transition pattern clearer.

Keep going

Read next

Continue with a related concept, model, or example.

Theory model

IS-LM Model

a well-connected theoretical model two steps from here

New angle: theoretical model →
Macro by Mark
Help
Privacy PolicyTerms of UseEthics and Compliance

© 2026 Mark Jayson Nation LLC