maeforecast: Economic Forecasting with High-Dimensional Data
In google-trends-v1/gtm: Functions for Economic Nowcasting with Google Trends Data

Description Usage Arguments Details Value Note See Also

This function calculates one-step-ahead point forecasts with recursive, rolling or fixed windows in a data-rich environment. Supported machine learning algorithms include Lasso, Ridge, Adaptive Lasso, Adaptive Elastic Net, as well as AR(1) model as a potential benchmark. Dynamic factor models, Regression Trees and Random Forest algorithms are also supported. Out-of-sample forecasts are returned alongside mean squred errors and success ratios.

maeforecast(data, model="ar", w_size, window="recursive", y.index=1,
            h=0, ...)

maeforecast.ar(data, w_size, window="recursive", y.index=1, h=0)
maeforecast.rw(data, w_size, windoe="recursive", y.index=1, h=0)
maeforecast.lasso(data, w_size, window="recursive", y.index=1,
            standardize=TRUE, lambda, h=0, t.select, t.update=F)
maeforecast.postlasso(data, w_size, window="recursive", y.index=1,
            standardize=TRUE, lambda, h=0, t.select, t.update=F)
maeforecast.ridge(data, w_size, window="recursive", y.index=1,
            standardize=TRUE, lambda, h=0, t.select, t.update=F)
maeforecast.alasso(data, w_size, window="recursive", y.index=1,
            standardize=TRUE, lambda.ridge, lambda.lasso, h=0,
            t.select, t.update=F)
maeforecast.postalasso(data, w_size, window="recursive", y.index=1,
            standardize=TRUE, lambda.ridge, lambda.lasso, h=0,
            t.select, t.update=F)
maeforecast.postnet(data, w_size, window="recursive", pred, y.index=1,
            standardize=TRUE, h=0, t.select, t.update=F,
            alphas=c(0.2, 0.8, 0.02))
maeforecast.rf(data, w_size, window="recursive", ntree=500, y.index=1,
            replace=TRUE, h=0, t.select, t.update=F)
maeforecast.dfm(data, w_size, window="recursive", y.index=1,
            factor.num, h=0, t.select, t.update=F)
maeforecast.dfm2(data, w_size, window="recursive", y.index=1,
            factor.num=3, method="two-step", clustor.type="partitional",
            h=0, t.select, t.update=F)

`data`	a data frame or a matrix containing the time series for which the forecasts are made as well as predictors.
`model`	character, indicating which model should be used to make the forecasts. Default is an AR(1) model. Note that `maeforecast(model="ar")` is equivalent to `maeforecast.ar()` and so on. See Details for the full list of supported models.
`w_size`	numeric, indicating the index where the forecasting should begin. If the first point forecast should be made at the 73th observation, for example, `w_size` should be set to be 72.
`window`	character, indicating the forecasting scheme to be applied. Options include `"recursive"`, `"rolling"`, and `"fixed"`.
`y.index`	numeric, indicating the column position of the time series for which the forecasts are made (Y). Defualt is `1`.
`pred`	numeric, indicating the number of predicators being considered in the Adaptive Elastic Net model. Default is set to be equal to the number of observations.
`alphas`	vector of candidate `alpha`'s to use in `aenet`.
`factor.num`	numeric, indicating the number of dynamic factors to be extracted from the covariates in the Dynamic Factor Model. Default is `3`.
`standardize`	logical, indicating whether the data matrix should be scaled before the model is fitting, for the use of variable selection/shrinkage models. Default is `TRUE`.
`ntree`	numeric, number of trees to grow. Default is `500`.
`replace`	logical, indicating whether sampling of cases should be done with replacement. Default is `TRUE`.
`method`	character, indicating which method should be used to extract dynamic factors. If `"aggregation"`, the covariate time series are first clustered based on partitional method, and a simple aggreagtion is applied to each cluster. If `"two-step"`, one factor is extracted within each cluster based on Doz, Giannone & Reichlin (2011).
`lambda, lambda.ridge, lambda.lasso`	optional user-supplied lambda sequence; default is NULL, and the function `glmnet` chooses its own sequence.
`h`	forecasting horizon. Default is `0`.
`t.select`	number of covariates to be included. If omitted, every covariate will be included. Otherwise, a regression between the dependant variable, its lag and each covariate will be run and a statistical test will be applied for the significance of the covariate's coefficient. The covariates will then be ranked based on their test statistics, and `t.select` number of them with the highest test statistics will be included in the machine learning algorithms. Note that the forecasting horizon `h` is considered.
`t.update`	logical, indicating wheter the preselection process should be repeated in evert iteration, if `t.select` is specified. Under the defualt `FALSE`, the preselection process will be implemented only with for the first window.

Supported models include "ar", "rw" (Random Walk), "lasso", "postlasso" (Post-Lasso), "ridge", "alasso" (Adaptive Lasso), "postalasso" (Post-AdaptiveLasso), "postnet" (Post-Adaptive ElasticNet), "rf" (Random Forests), "dfm" & "dfm2" (Dynamic Factor Models).

For shrinkage models, the penalty factor lambda is selected automatically by 10-fold cross-validation.

`Forecasts`	data matrix, containing the point forecasts, realized values, forecast errors, signs of the forecasts and realized values, and success in predicting the signs.
`MSE`	numeric, mean squred error of the point forecasts.
`SRatio`	numeric, success ratio of the point forecasts. Success is claimed when the point forecasts and realized values have the same sign.
`Data`	the data as used in the model.
`Model`	some specifics about the model used.