How to Implement Some Multiple Imputation Models for Panel Data

EasyChair Preprint 9007

Abstract

Missing data is a challenge in statistical analyzes because the results they yield have limitations. Imputation, understood as the process of replacing missing data with an estimated value, is a regular problem in research projects. There are many models and packages intended for this process, however, the selection of the appropriate imputation model for the type of data available is crucial for the reliability of the result. This study works with a cross-data table involving time series (panel data) for 33 countries and 17 variables (annual Gini Index for the period 2000-2016), with 24% missing data. In order to impute these data, a multiple imputation model proposed by Honaker and King (2010) was used and some restrictions were added to the system. The main results obtained lead to the following question: Can the imputation be trusted? All the files needed to reproduce the presented results are available at: \url{https://gitlab.com/iesta.fcea.udelar/how-to-implement-some-multiple-imputation-models-for-data- pane} .

Keyphrases: Datos faltantes, Datos panel, Imputación

@booklet{EasyChair:9007,
  author    = {Ramón Álvarez-Vaz and Diana Del-Callejo-Canal and Margarita Edith Canal-Martínez and Elena Vernazza and Alar Urruticoechea},
  title     = {How to Implement Some Multiple Imputation Models for Panel Data},
  howpublished = {EasyChair Preprint 9007},
  year      = {EasyChair, 2022}}