Diagnostics for detecting low variability Public Deposited

Analytics

Downloadable Content

Download PDF
Request Version for Screen Reader
Last Modified
  • March 22, 2019
Creator
  • Arizmendi, Cara
    • Affiliation: College of Arts and Sciences, Department of Psychology and Neuroscience
Abstract
  • Whether to include variables with low variability in analyses is a predicament for researchers. Like the case of multicollinearity in which a correlation matrix has a near-zero determinant, data with low-variability variables will have a near-zero determinant of the covariance matrix. Similar to problems with multicollinearity, we may expect that retaining low-variability variables in our models would result in unstable estimates of the other variables in the model. Despite this predicament: 1. a method for identifying these variables for removal has yet to be developed and 2. the impact of low variability on estimation of other variables has yet to be explored. We present a simulation study exploring the impact of low variability in linear regression, while also assessing indices of low variability. Simulations suggest the stability of estimates of normal-variability variables, despite the presence of a low variability variable and the instability of estimates for variables with low variability.
Date of publication
Keyword
DOI
Resource type
Advisor
  • Thissen, David
  • Gates, Kathleen
  • Bollen, Kenneth
Degree
  • Master of Arts
Degree granting institution
  • University of North Carolina at Chapel Hill Graduate School
Graduation year
  • 2018
Language
Parents:

This work has no parents.

Items

Thumbnail Title Date Uploaded Visibility Actions
Arizmendi_unc_0153M_17679.pdf 2019-04-12 Public
PREMIS_Events_Metadata_0_ca596dc1-6ab1-47be-aaee-6cf3e7540b8d.txt 2019-04-12 Public
original_metadata_file_ca596dc1-6ab1-47be-aaee-6cf3e7540b8d.xml 2019-04-12 Public