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

Comparisons of two-part models with competitors.

P A Lachenbruch1

  • 1FDA/CBER Division of Biostatistics and Epidemiology, Food and Drug Administration, 1401 Rockville Pike, Rockville, MD 20852-1448, USA. lachenbruch@cber.fda.gov

Statistics in Medicine
|April 17, 2001
PubMed
Summary

This study compares statistical tests for two-part models with zero-inflated data. Two-part models with chi-squared tests perform best when higher zero proportions align with larger means.

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

Observations on the relationship between frequency and timing of intercourse and the probability of conception.

Population studies·2011
Same author

Frequency and timing of intercourse: Its relation to the probability of conception.

Population studies·2011
Same author

The Cutaneous Assessment Tool: development and reliability in juvenile idiopathic inflammatory myopathy.

Rheumatology (Oxford, England)·2007
Same author

Lot consistency as an equivalence problem.

Journal of biopharmaceutical statistics·2004
Same author

Repeated measures with zeros.

Statistical methods in medical research·2002
Same author

Analysis of studies to evaluate immune response to combination vaccines.

Clinical infectious diseases : an official publication of the Infectious Diseases Society of America·2001

Area of Science:

  • Biostatistics
  • Statistical Modeling
  • Data Analysis

Background:

  • Two-part models are essential for analyzing non-negative continuous data with a concentration of zero values.
  • Comparing distributions with excess zeros requires specialized statistical approaches beyond standard tests.

Purpose of the Study:

  • To compare the statistical power and accuracy (size) of various methods for comparing two-part distributions.
  • To evaluate the performance of different statistical tests under varying conditions of zero proportions and mean differences.

Main Methods:

  • Comparison of established statistical tests: z-test (or t-test), Wilcoxon-Mann-Whitney rank sum test, and Kolmogorov-Smirnov test.
  • Evaluation of two-part models utilizing a 2 degree of freedom chi-squared test, combining proportion and continuous part tests.

Related Experiment Videos

  • Assessment of test performance under the null hypothesis and analysis of power for detecting alternatives, considering finite moments.

    • Main Results:

    • The 2 degree of freedom chi-squared tests demonstrate superior performance when a higher proportion of zeros corresponds to a larger mean.
    • When the proportion of zeros is inversely related to the mean, single-part models (rank sum or Kolmogorov-Smirnov) show the best performance.
    • Two-part models maintain competitive power, though statistically less powerful than the best single-part models in specific scenarios.

    Conclusions:

    • The choice of statistical test for two-part models depends critically on the relationship between the proportion of zeros and the mean.
    • The z-test lacks power for distributions without finite moments, highlighting the importance of model assumptions.
    • Two-part models offer a robust framework, but their effectiveness is nuanced by the specific data characteristics and chosen comparison tests.