An approach for estimating twin similarity for the case of count data is proposed. The
methodology focuses on the Intracluster Correlation Coefficient (ICC) as the parameter of
interest. This is in analogy to the methods commonly applied for continuous random variables.
The ICC is related to the parameters of the Negative Binomial distribution. Maximum
Likelihood Estimation of the ICC's from two samples representing the two zygosity groups and
their difference is considered to study twin similarity. Point and Interval estimation are
considered, as well as testing of hypothesis.