Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Videos

When can group level clustering be ignored? Multilevel models versus single-level models with sparse data.

P Clarke1

  • 1Institute for Social Research, University of Michigan, 426 Thompson Street, Ann Arbor, MI 48106-1248, USA. pjclarke@umich.edu

Journal of Epidemiology and Community Health
|July 16, 2008
PubMed
Summary
This summary is machine-generated.

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Relieving veterinarians' workloads and stress: leveraging Australia's veterinary technologists and nurses.

Australian veterinary journal·2023
Same author

Exploring the impact of athletic identity on gender role conflict and athlete injury fear avoidance in male English professional academy football players.

Science & medicine in football·2023
Same author

Scintillation light detection in the 6-m drift-length ProtoDUNE Dual Phase liquid argon TPC.

The European physical journal. C, Particles and fields·2022
Same author

GM-CSF Promotes Immune Response and Survival in a Mouse Model of COVID-19.

Research square·2022
Same author

Prospects for beyond the Standard Model physics searches at the Deep Underground Neutrino Experiment: DUNE Collaboration.

The European physical journal. C, Particles and fields·2021
Same author

Personalising screening of sight-threatening diabetic retinopathy - qualitative evidence to inform effective implementation.

BMC public health·2020
Same journal

Sociodemographic inequalities in non-small cell lung cancer (NSCLC) novel treatments: a national population-based cohort study.

Journal of epidemiology and community health·2026
Same journal

Secular trends in primary care utilisation for mental health problems: a Norwegian register-based population-wide study.

Journal of epidemiology and community health·2026
Same journal

Guide to recent advances in difference-in-differences methodology for population health studies.

Journal of epidemiology and community health·2026
Same journal

From early to fewer first births: ADHD and family formation among young adults.

Journal of epidemiology and community health·2026
Same journal

10-year mortality among first-time mothers involved in family court care proceedings in England: cohort study using linked administrative hospital, mortality and family court records.

Journal of epidemiology and community health·2026
Same journal

Intersectional inequalities in self-rated health in Germany across sex/gender, migration, education and income: a cross-sectional study.

Journal of epidemiology and community health·2026
See all related articles

Multilevel modeling is reliable with five observations per group. Single-level regression risks Type I errors, especially with discrete outcomes, even with sparse data.

Area of Science:

  • Statistical modeling
  • Survey methodology
  • Biostatistics

Background:

  • Multilevel modeling is often limited by small numbers of cases per level-2 unit in population-based surveys.
  • Researchers frequently resort to single-level techniques like ordinary least squares regression due to data sparseness.
  • This limitation impacts the validity of complex statistical analyses in various research fields.

Purpose of the Study:

  • To investigate the effects of data sparseness on parameter estimates in two-level versus single-level models.
  • To compare the performance of multilevel models with disaggregated techniques under conditions of limited data.
  • To examine differences in the impact of small group size across continuous and discrete outcomes.

Main Methods:

  • Monte Carlo simulations were employed to assess parameter estimate validity.

Related Experiment Videos

  • Both linear and non-linear hierarchical models were simulated.
  • Simulated results were compared against ordinary least squares and logistic regression.

    • Main Results:

    • Two-level models overestimated group-level variance with extreme sparseness (two observations/group).
    • Valid estimates were achieved with an average of five observations/group for both continuous and discrete outcomes.
    • Single-level models risked Type I errors (downward biased standard errors), particularly with discrete outcomes.

    Conclusions:

    • Multilevel models demonstrate reliable estimation capabilities with an average of five observations per group.
    • Disaggregated techniques increase the risk of Type I errors, even with minimal data clustering.
    • These findings support the use of multilevel modeling in population-based surveys with limited group sizes.