2

Suppose there are three groups of subjects: patients (P), controls (C), and relatives of patients (R). The investigator would like to conduct group comparisons on a behavioral measure. The complexity here is that some of the subjects in group R are genetically related within the R group as well as with some subjects in group P. Furthermore, the genetic relatedness varies in the following fashion:

50% genetic relatedness between a parent and child

50% between two full siblings

25% between two half siblings

The comparison between groups P and C is straightforward (conventional two-sample t-test), but the trouble is for the following two group comparisons: (1) between groups C and R, and (2) between groups P and R.

Any suggestions for handling the genetic relatedness when performing these two comparisons?

bluepole
  • 2,699
  • It seems like a problem for hierarchical Bayesian model with hyperpriors for parameters of related subjects. – Tim Jan 08 '15 at 19:58
  • @Tim: Thanks for suggesting hierarchical Bayesian model. Do you happen to know any implementation of that approach, for example, in R? – bluepole Jan 08 '15 at 21:52
  • 1
    The most up-to-date are Stan and JAGS however notice that this needs a little bit of learning on both this software and theory. But I would consider this approach as the best in here since it enables you to directly model the probabilistic relations between your data. – Tim Jan 09 '15 at 07:23

1 Answers1

1

Is your goal to assess the heritability of a certain phenotype? You will probably want to use a structural equation modeling approach. In doing this, it's possible for you to specify relatedness in that fashion. In particular you may want to consider what are called ACE models where the A stands for additive genetic contributions, C stands for common environmental (upbringing) contributions, and E stands for environmental contributions. Rotations of phenotypic variations are broke down to estimate the percentage of variability in phenotypes attributable to these factors using any combinations of these features.

AdamO
  • 62,637
  • Thanks for the suggestion. No, my goal is not about the heritability. Instead, I'm only interested in the group comparisons while properly accounting for the genetic relatedness among some subjects. – bluepole Jan 08 '15 at 20:05