Skip to contents

Getting Started with bridgr

Source: vignettes/bridgr.Rmd
bridgr.Rmd

Overview

bridgr is a mixed-frequency forecasting package for bridge models, MIDAS-style regressions, and intermediate specifications that estimate within-period weights from the data.

The core workflow is always the same:

  1. Provide one lower-frequency target series and one or more higher-frequency indicators.
  2. Decide how missing indicator observations should be handled with indic_predict.
  3. Decide how indicators should be aligned to the target frequency with indic_aggregators.
  4. Fit the target equation with bridge(), inspect it with summary(), and produce target-period forecasts with forecast().

This vignette walks through that workflow with the package’s built-in Swiss GDP and barometer data.

Example Data 

gdp_growth <- suppressMessages(tsbox::ts_na_omit(tsbox::ts_pc(gdp)))

head(gdp_growth)
#> # A tibble: 6 × 2
#>   time       values
#>   <date>      <dbl>
#> 1 2004-04-01  0.839
#> 2 2004-07-01 -0.104
#> 3 2004-10-01  0.242
#> 4 2005-01-01  0.860
#> 5 2005-04-01  1.06 
#> 6 2005-07-01  1.15
head(baro)
#> # A tibble: 6 × 2
#>   time       values
#>   <date>      <dbl>
#> 1 2004-01-01   109.
#> 2 2004-02-01   108.
#> 3 2004-03-01   109.
#> 4 2004-04-01   110.
#> 5 2004-05-01   109.
#> 6 2004-06-01   105.

gdp_growth is quarterly, while baro is monthly. bridge() recognizes the frequency mismatch automatically and aligns the indicator to the target frequency before fitting the target equation.

A Basic Bridge Model 

bridge_model <- bridge(
  target = gdp_growth,
  indic = baro,
  indic_predict = "auto.arima",
  indic_aggregators = "mean",
  indic_lags = 1,
  target_lags = 1,
  h = 2
)

forecast(bridge_model)
#> Bridge forecast
#> -----------------------------------
#> Target series: gdp_growth
#> Forecast horizon: 2
#> Uncertainty: point forecast only
#> -----------------------------------
#>   time       mean 
#> 1 2023-01-01 0.972
#> 2 2023-04-01 0.698

The default mean aggregator is the classic bridge-model setup: each monthly block is completed first, then averaged to the quarterly frequency before the target equation is estimated.

The fitted object stores the aligned data that went into estimation and the future target-period regressor path used for forecasting.

tail(bridge_model$estimation_set)
#> # A tibble: 6 × 4
#>   time       gdp_growth  baro baro_lag1
#>   <date>          <dbl> <dbl>     <dbl>
#> 1 2021-07-01      2.34  112.      125. 
#> 2 2021-10-01      0.411 104.      112. 
#> 3 2022-01-01      0.105  97.4     104. 
#> 4 2022-04-01      1.03   94.3      97.4
#> 5 2022-07-01      0.255  90.0      94.3
#> 6 2022-10-01      0.102  90.7      90.0
bridge_model$forecast_set
#> # A tibble: 2 × 3
#>   time        baro baro_lag1
#>   <date>     <dbl>     <dbl>
#> 1 2023-01-01  97.4      90.7
#> 2 2023-04-01  99.8      97.4

Standardized Output

summary() and forecast() use a stable package-specific layout. The base output is the same across bridge, MIDAS-style, and direct-alignment specifications. Additional details, such as optimization summaries or bootstrap settings, are appended only when they are relevant.

summary(bridge_model)
#> Bridge model summary
#> -----------------------------------
#> Target series: gdp_growth
#> Target frequency: quarter
#> Forecast horizon: 2
#> Estimation rows: 74
#> Regressors: baro, baro_lag1
#> -----------------------------------
#> Target equation coefficients:
#>           Estimate
#> ar1          0.243
#> intercept   -6.493
#> baro         0.161
#> baro_lag1   -0.092
#> -----------------------------------
#> Indicator summary:
#>      Frequency Predict    Aggregation
#> baro month     auto.arima mean       
#> -----------------------------------

Forecast Visualization

The package also provides a built-in plotting method for fitted bridge models. With type = "forecast", it shows the observed target history together with the forecast path generated by the model.

plot(bridge_model, type = "forecast")

Direct Alignment

If you set indic_predict = "direct", bridgr switches from indicator forecasting to direct MIDAS-style alignment. In that case, the latest complete high-frequency blocks are assigned backward to target periods instead of being forecast forward first.

direct_model <- bridge(
  target = gdp_growth,
  indic = baro,
  indic_predict = "direct",
  indic_aggregators = "unrestricted",
  h = 1
)

forecast(direct_model)
#> Bridge forecast
#> -----------------------------------
#> Target series: gdp_growth
#> Forecast horizon: 1
#> Uncertainty: point forecast only
#> -----------------------------------
#>   time       mean 
#> 1 2023-01-01 0.483

This is particularly useful at the ragged edge when you want to work only with observed high-frequency information and avoid a separate indicator forecasting step.

Optional Bootstrap Uncertainty

By default, bridgr returns point forecasts only. If you want uncertainty output, request it at estimation time with se = TRUE and a bootstrap configuration.

boot_model <- bridge(
  target = gdp_growth,
  indic = baro,
  indic_predict = "auto.arima",
  indic_aggregators = "mean",
  target_lags = 1,
  h = 4,
  se = TRUE,
  bootstrap = list(type = "block", N = 40, block_length = NULL)
)

forecast(boot_model)
#> Bridge forecast
#> -----------------------------------
#> Target series: gdp_growth
#> Forecast horizon: 4
#> Uncertainty: predictive intervals from conditional block bootstrap
#> Bootstrap draws: 40 / 40
#> Block length: 5
#> -----------------------------------
#>   time       mean  se    lower_80 upper_80 lower_95 upper_95
#> 1 2023-01-01 0.116 0.938 -1.191   1.104    -1.790   2.586   
#> 2 2023-04-01 0.428 0.920 -0.920   1.496    -2.233   1.834   
#> 3 2023-07-01 0.487 0.994 -0.631   2.207    -1.324   3.017   
#> 4 2023-10-01 0.509 1.104 -0.504   2.036    -2.696   2.799
summary(boot_model)
#> Bridge model summary
#> -----------------------------------
#> Target series: gdp_growth
#> Target frequency: quarter
#> Forecast horizon: 4
#> Estimation rows: 75
#> Regressors: baro
#> -----------------------------------
#> Target equation coefficients:
#>           Estimate Bootstrap SE
#> ar1         -0.135        0.203
#> intercept   -9.060        3.626
#> baro         0.095        0.036
#> -----------------------------------
#> Indicator summary:
#>      Frequency Predict    Aggregation
#> baro month     auto.arima mean       
#> -----------------------------------
#> Uncertainty:
#> Method: conditional block bootstrap with predictive forecast draws
#> Bootstrap draws: 40 / 40
#> Block length: 5
#> -----------------------------------
plot(boot_model, type = "forecast")

The current uncertainty implementation uses a conditional block bootstrap on the final target-frequency estimation sample. It does not re-estimate indicator forecast models or parametric aggregation weights inside each bootstrap draw. The stored forecast draws do add simulated future target shocks, so forecast() and plot(..., type = "forecast") return predictive intervals on top of the historical target path and forecast mean.

Where to Go Next

The vignette vignette("mixed-frequency-modeling", package = "bridgr") compares the main aggregation strategies and shows how bridgr moves from classic bridge models to unrestricted and parametric MIDAS-style specifications.

The vignette vignette("ragged-edge-nowcasting", package = "bridgr") focuses on indic_predict and the different ways to handle incomplete high-frequency data at the forecast origin.

The vignette vignette("uncertainty-and-scenarios", package = "bridgr") shows how to work with conditional block-bootstrap uncertainty and scenario forecasts based on custom future regressor paths.