Consider an example (like the Georgia birthweight example) with before and after measurements on a subject. If the variability of the before and after measurements each have variance σ 2 and...

Consider an example (like the Georgia birthweight example) with before and after measurements on a subject. If the variability of the before and after measurements each have variance σ²
and correlation ρ then it is a fact that the standard deviation of the difference is

(1) The correlation of the first and last birthweights is about 0.381. Using Table 7.5, verify the above formula (approximately).

(2) If we were to compare two groups, based on the difference scores or just the last birthweights (say, those with initial age greater than 17 versus those not), which analysis would have a larger variance and hence be less powerful? By how much?