Post date: Jan 04, 2014 10:21:39 PM
Regarding GLA on Ms, when dropping 20% of individuals for cross-validation the mean latent survival score for survivors was 0.568 vs 0.448 (difference = 0.121). This is significantly greater than expected by chance (permutation test, 10,000 permutations, p = 0.0019). Whereas there is a trend in the same direction for cross-validation based on 20% of families (survivors = 0.510, dead = 0.456, difference = 0.053), it is not significantly greater than expected by chance (p = 0.1185, though this is low enough to be suggestive). This weaker signal indicates that (i) much of the predictive information comes from the polygenic effect which requires closely related kin to infer, (ii) the predictive information comes in part from non-genetic maternal effects, or (iii) both. Nonetheless, there is enough signal here to make the model interesting.