State-space Model Multiple Imputation for Missing Data in non-stationary Multivariate Time Series • SSMimpute

The goal of SSMimpute is to address missing data in possibly non-stationary multivariate time series. This is a multiple imputation strategy based on the state space model. This package implements the methods SSMimpute introduced by Xiaoxuan Cai et al., 2022.

Installation

You can install the development version of SSMimpute like so:

devtools::install_github("js5959/SSMimpute")
library(SSMimpute)

Work Flow

Data exploration: Using shrinkTVP() and explore_SSM() function to get some sense of the data, identify the formula, check the possible ARIMA model of latent space and find the change points. The problem of over fitting is partially alleviated by introducing shrinkage. Based on the exploration result, we set up the ss_param and cpt_learning_param for the SSMimpute_unanimous_cpts() function. See Introduction to Exploration for detail process of exploration.
Once you set up the parameters based on the exploration, run SSMimpute_unanimous_cpts(). You will have estimated coefficient for state-space model, and the imputed missing data. See Quick Start for a detailed illustration of function usage and analysis of results.