Robustness of path analysis of family resemblance against deviations from multivariate normality

D. C. Rao, G. P. Vogler, I. B. Borecki, M. A. Province, J. M. Russell

Research output: Contribution to journalArticlepeer-review

7 Scopus citations

Abstract

Path analysis is one of several methods available for quantitative genetic analysis, providing for both tests of hypotheses and estimates of relevant parameters. Central to the theory is the assumption that the observations follow a multivariate normal distribution within families. The purpose of the present investigation is to assess the effects of a certain type of departures from multivariate normality using quantitative family data on lipid and lipoprotein levels. The results show that even large departures produce reasonably unbiased parameter estimates. Whereas moderate departures lead to few inferential errors in hypothesis testing, gross departures from multivariate normality may have considerable effects on likelihood ratio tests.

Original languageEnglish
Pages (from-to)107-112
Number of pages6
JournalHuman heredity
Volume37
Issue number2
DOIs
StatePublished - 1987

Keywords

  • Inference
  • Multivariate normality
  • Parameter estimates
  • Path analysis

Fingerprint

Dive into the research topics of 'Robustness of path analysis of family resemblance against deviations from multivariate normality'. Together they form a unique fingerprint.

Cite this