• A
  • A
  • A
  • ABC
  • ABC
  • ABC
  • А
  • А
  • А
  • А
  • А
Regular version of the site

Working paper

A Comparison of the Missing-Indicator Method and Complete Case Analysis in Case of Categorical Data

Sociology. WP BRP. National Research University Higher School of Economics, 2019
The research aims to provide a complex analysis of the missing-indicator method’s performance in case of a categorical independent variable in regression in comparison with complete case analysis. While the latter seems to be the most popular way to handle missing data, the former appears to be a simple and effective alternative that allows making a full sample available for analysis. By means of a statistical experiment and simulated data, we examined how these methods perform in conditions that differ in a mechanism of missingness, proportion of missing data, and model specification. The final results show that, overall, both methods produce unbiased estimates of regression coefficients, but crucially biased estimates of their standard errors and additional statistics such as R2, adjusted R2, and F-statistic, especially in case of a missing-indicator method. We explain these results by contribution of a missing-indicator variable, coefficient of which always turns out to be significant and far away from zero.