gen_atomic_preds: Generate atomic predictions

In ooelrich/oscbvar: Stuff for Bayesian vector autoregressions

Description

Generates a notebook of atomic predictive densities and means.

Usage

gen_atomic_preds(model_list, agc = list(5, 60, FALSE, 1))

Arguments

model_list

List of names of the atomic models to use.

agc

List of atomic prediction generation controllers with four elements: Which observation to consider as the first observation (ie t=1). Defaults to 5, which gives some breathing room for the specification of AR models. Size of the estimation window. This determines how many observations are used to train the models before any predictions are made. Defaults to 60. Whether to use a rolling estimation window or not. The default is FALSE, which corresponds to a non-rolling window. Which variable should be the response variable. Defaults to 3 (FEDFUNDS). Other possible values are 1 (GDP) and 2 (GDPTCPI).

Details

Function to generate the atomic predictions (that is the predictions of the individual models, not of aggregation schemes) for the macrodata example. Essentialy a list of potential models to include that the user selects from. Also contains some global settings that always should be the same between models, such as estimation window length, starting time, and which data set to use.

Value

A data frame of atomic predictions with columns

1. pmean, the predictive mean for time t (made at t-1)

2. lpdens, the log predictive density for time t (made at t - 1)

3. method, which model made the prediction

4. t, timepoint

5. ytrue, true value of the response at time t

ooelrich/oscbvar documentation built on Sept. 8, 2021, 3:31 p.m.