A latent class model for bivariate binary responses from twins

Ekholm, Anders, Jokinen, Jukka and Smith, Peter (2006) A latent class model for bivariate binary responses from twins , Southampton, UK University of Southampton, Southampton Statistical Sciences Research Institute 23pp. (S3RI Methodology Working Papers, M06/10).


[img] PDF 39276-01.pdf - Author's Original
Download (448kB)


We analyse a 2^4 table reporting the presence or absence of alcohol dependence and depression in both members of 597 pairs of female monozygotic twins. The statistical analysis is based on a latent class model, formulated as a 2x2 table and parametrized by two marginal univariate genetic dispositions and the dependence ratio between these dispositions. The final model, selected after some empirically motivated simplifying assumptions are adopted, has six parameters and fits the data very well. The identifiability and stability of the final model is studied by contour plots of profile log-likelihood functions with two arguments. Numerical results are compared to those obtained when the measure of association between binary responses is the odds ratio rather than the dependence ratio. The conditional probability of alcohol dependence given the disposition for it is approximately 1/3, while the corresponding probability of depression is approximately 2/3. The association between these two latent dispositions is strong, the dependence ratio being approximately 3. An extended model, defining measures of heritability and similarity, for analysing, simultaneously, data on monzygotic and dizygotic twins is proposed.

Item Type: Monograph (Working Paper)
Keywords: contour plot, dependence and odds ratio, dizygotic and monozygotic twin, gebetic disposition, cohen's kappa, moment parameterization

ePrint ID: 39276
Date :
Date Event
Date Deposited: 26 Jun 2006
Last Modified: 16 Apr 2017 21:56
Further Information:Google Scholar
URI: http://eprints.soton.ac.uk/id/eprint/39276

Actions (login required)

View Item View Item