1 The sample

In this section, we apply sex-limitation models to data on body mass index collected from twins in the Virginia Twin Registry and twins ascertained through the American Association of Retired Persons (AARP). Details of the membership of these two twin cohorts are provided in Eaves et al. (1991), in their analysis of BMI in extended twin-family pedigrees. In brief, the Virginia twins are members of a population based registry comprised of 7,458 individuals (Corey et al., 1986), while the AARP twins are members of a volunteer registry of 12,118 individuals responding to advertisements in publications of the AARP. The Virginia twins' mean age is 39.7 years (SD = 14.3), compared to 54.5 years (SD = 16.8) for the AARP twins. Between 1985 and 1987, Health and Lifestyle questionnaires were mailed to twins from both of these cohorts. Among the items on the questionnaire were those pertaining to physical similarity and confusion in recognition by others (used to diagnose zygosity) and those asking about current height and weight (used to compute body mass indices). Questionnaires with no missing values for any of these items were returned by 5,465 Virginia and AARP twin pairs.

From height and weight data, body mass indices (BMI) were calculated for the twins, using the formula:

$$\mbox{BMI} = ht (m) ^{2}/wt (kg)$$

The natural logarithm of BMI was then taken to normalize the data. Before calculating covariance matrices of log BMI, the data from the two cohorts were combined, and the effects of age, age squared, sample (AARP vs. Virginia), sex, and their interactions were removed. The resulting covariance matrices are provided in the Mx scripts in Appendices  and , while the correlations and sample sizes appear in Table 9.1 below.

Table 9.1: Sample sizes and correlations for BMI twin data.

Zygosity Group	N	r
MZF	1802	0.744
DZF	1142	0.352
MZM	750	0.700
DZM	553	0.309
DZO	1341	0.251

We note that both like-sex MZ correlations are greater than twice the respective DZ correlations; thus, models with dominant genetic effects, rather than common environmental effects, were fit to the data.