A proposed test of the difference between the means of monozygotic and dizygotic twins

The rationale for using an approximate t' test of the difference between the means for MZ and DZ twins is presented. This test avoids pooling the among-MZ and among-DZ twin-pair mean squares and has approximate degrees of freedom based upon the relative number of MZ and DZ twin pairs as well as the relative sizes of the among-MZ and among-DZ twin-pair mean squares. Sampling experiments simulating twin studies were used to show that the rate of Type I error for this t' test was appropriate while other tests could give either too many or too few Type I errors depending upon the relative sizes of the mean squares and the relative numbers of MZ and DZ twin pairs.