Dear all,
Further to the recent queries about Cramer's V and Tshuprow's T
coefficients for contingency table, which are related to the expectation of
chi-squared statistic and available from statistical packages such as SAS
(but does not report standard errors) and SPSS. We use such summary
statistics for allelic association between genetic markers.
In the classic reference "Discrete Multivariate Analysis -- Theory and
Practice" by Bishop, Y. M. M., S. E. Fienberg, et al. (1975) , The MIT
press, they also discussed Kullback-Liebler information measure but it is
not directly available from packages such as SAS. Intuitively it can be
extended to multidimensional table but may suffer from sparseness of the
table. A likelihood-based measure has therefore been proposed based on
permutation, e.g., Zhao, H., A. J. Pakstis, et al. (1999). "Assessing
linkage disequilibrium in a complex genetic system. I. Overall deviation
from random association." Annals of Human Genetics 63: 167-179.
Therefore I would like to hear some general comments about these.
Many thanks in advance,
Jing Hua Zhao
Institute of Psychiatry
See 2LD
http://www.iop.kcl.ac.uk/IoP/Departments/PsychMed/GEpiBSt/software.stm
for such a program
|