fbi: Factor-Based Imputation for Missing Data

The codes in the fbi package are developed based on the following papers:

Bai, Jushan and Ng, Serena (2021), "Matrix Completion, Counterfactuals, and Factor Analysis of Missing Data".
Cahan, E., Bai, J. and Ng, S. (2021), "Factor-Based Imputation of Missing Values and Covariances in Panel Data of Large Dimensions".
Bai, Jushan and Ng, Serena (2002), "Determining the number of factors in approximate factor models".
Bai, Jushan and Ng, Serena (2019), "Rank regularized estimation of approximate factor models".
McCracken, Michael W. and Ng, Serena (2015), "FRED-MD and FRED-QD: Monthly and Quarterly Databases for Macroeconomic Research".

Requirements

The fbi package requires the following three R packages:

stats
readr
pracma

They should be installed prior to the installation of the fbi package:

install.packages("stats")
install.packages("readr")
install.packages("pracma")

Installation

The fbi package can be installed from github:

devtools::install_github("cykbennie/fbi")

Reference Manual

Refer to the fbi.pdf file for details.

Vignette

Download factor_fred.html for an example using the FRED-MD dataset (https://research.stlouisfed.org/econ/mccracken/fred-databases/).

Author

Yankang (Bennie) Chen ([email protected])

License

This project is licensed under the GNU General Public License v3.0 License - see the LICENSE file for details.

Acknowledgments

Serena Ng ([email protected])
Jushan Bai ([email protected])

fbi's People

Contributors

Stargazers

Watchers

fbi's Issues

Error in fredqd() while retrieving the dataset using the current civ file

To replicate:

fl <- "https://files.stlouisfed.org/files/htdocs/fred-md/quarterly/current.csv" df <- fredqd(file = fl, transform = TRUE)

It returns the following error:

Error in 1:ind_notna : result would be too long a vector
In addition: Warning message:
In min(which(is.na(rawdata[, 1]))) :
no non-missing arguments to min; returning Inf

fredmd_description does not contain recent changes

Is the data set fredmd_description updated?
It seems it doesn't not contain the recent changes, e.g. for VXOCLSx changed to VIXCLSx.

missing most recent observations

For most vintages the most recent month seems to be missing when loading (downloaded) fredmd csv files with the fred_md() function. For instance when loading the first vintage (1999-08.csv), the last available month is June whereas some observations are available for July in the csv file.

After looking in the code I think it is because of line 51 in fredmd.R : rawdata <- rawdata[1:(nrow(rawdata) - 1), ] # remove NA rows
I think the last raw should be removed only if it contains only missing values

Recommend Projects

cykbennie / fbi Goto Github PK

fbi's Introduction