Brief Research Report: A Monte Carlo Simulation Study of Small Sample Bias in Ordered Logit Model under Multicollinearity

Eric Nimako Aidoo*, Simon K. Appiah, Alexander Boateng

*Corresponding author for this work

Research output: Contribution to journalArticlepeer-review

3 Citations (Scopus)

Abstract

This study investigated the small sample biasness of the ordered logit model parameters under multicollinearity using Monte Carlo simulation. The results showed that the level of biasness associated with the ordered logit model parameters consistently decreases for an increasing sample size while the distribution of the parameters becomes less variable with low extreme values. In the presence of multicollinearity, the level of biasness increases and this issue is particularly severe for small sample sizes. By comparing three different approaches for dealing with the multicollinearity problem in the model, the study demonstrated that the use of penalized maximum likelihood estimation technique provides better results which is interpretable compared to the other approaches considered.

Original languageEnglish
Pages (from-to)742-750
Number of pages9
JournalJournal of Experimental Education
Volume89
Issue number4
Early online date11 Jan 2020
DOIs
Publication statusPublished - 2 Oct 2021

Keywords

  • Multicollinearity
  • ordered logit model
  • penalized mle
  • principal component analysis
  • simulation
  • small sample

ASJC Scopus subject areas

  • Education
  • Developmental and Educational Psychology

Fingerprint

Dive into the research topics of 'Brief Research Report: A Monte Carlo Simulation Study of Small Sample Bias in Ordered Logit Model under Multicollinearity'. Together they form a unique fingerprint.

Cite this