Comparisons of offspring to controls in the total samples and in subsamples of different age periods.

1Comparison of total sample of offspring to total sample of controls.
– Non-independence of observations: The 65 offspring sample was composed of 26 singletons and 17 sibships: 12 comprised two subjects and 5 comprised three. To account for possible correlation among subjects within the same sibship, a multilevel model was carried out using the MIXED procedure of SAS (version 9.2; SAS Institute Inc., Cary, NC). Sibships nested in the group were used as the second level and modeled according to a random effect. Degrees of freedom were obtained by the method of Kenward-Roger, 1997 [46] that is available with the option DDFM = KR in the MODEL statement of the MIXED procedure.
Effect sizes (ES): ES were calculated using the difference of adjusted means (LSMeans) between the experimental and control groups standardized by a pooled standard deviation. The pooled standard deviation was obtained by dividing the standard error of the difference of LSMeans by the square root of , according to Kelley [47]. Confidence intervals (CI) for the effect sizes were obtained using the non-centrality interval estimation approach based on a t distribution, according to Steiger and Fouladi, 1997 [48]. The lower and upper bounds of the 95% CI were calculated by multiplying the 2.5% and 97.5% percentiles, respectively, of the non central t distribution by the square root of .
– Interaction term: A statistically significant interaction term (Group x Age Periods), as described in Statistical Analysis, was found only for the two Rey Complex Figures Tests (immediate recall and delayed recall, respectively RCFTIR and RCFTDR) suggesting that the cognitive differences for this domain varied with age. This imposed to interpret the effect sizes at each age period rather than the marginal one.