Skip to main content
. Author manuscript; available in PMC: 2020 Feb 20.
Published in final edited form as: Stat Med. 2019 Dec 4;39(4):438–455. doi: 10.1002/sim.8415

TABLE 4.

Percent relative bias of the correlation parameters based on uncorrected quasi-least squares (QLS) and matrix-adjusted QLS (MAQLS) for each simulation scenario when the treatment effect is zero

Correlations Effect size Design resource Percent bias for τ Percent bias for ρ
τ ρ δ I N T QLS MAQLS QLS MAQLS
0.03 0.2 0 18 10 7 −26.6 3.5 −0.6 0.2
0.03 0.2 0 18 24 4 −16.0 5.5 −0.3 0.0
0.03 0.2 0 20 14 5 −20.2 3.4 −0.5 0.0
0.03 0.2 0 21 8 4 −29.4 3.2 −0.4 0.2
0.03 0.2 0 15 8 4 −41.3 4.7 −0.6 0.3
0.03 0.8 0 16 12 5 −26.2 4.4 −0.6 0.0
0.03 0.8 0 24 7 5 −27.7 0.6 −0.6 −0.1
0.03 0.8 0 12 8 5 −49.4 8.2 −2.1 −0.4
0.03 0.8 0 12 5 4 −74.2 6.6 −2.3 −0.5
0.03 0.8 0 10 5 3 −91.6 −0.4 −1.5 −0.5
0.10 0.2 0 21 11 8 −9.2 3.9 −0.6 0.6
0.10 0.2 0 24 11 7 −8.1 3.4 −0.4 0.6
0.10 0.2 0 15 16 6 −11.8 7.1 −0.6 1.1
0.10 0.2 0 18 8 7 −12.7 3.9 −0.8 0.7
0.10 0.2 0 16 7 5 −16.7 3.7 −0.7 1.0
0.10 0.8 0 20 18 5 −8.0 5.0 −0.1 0.2
0.10 0.8 0 15 9 4 −15.2 3.1 −0.2 0.2
0.10 0.8 0 10 20 3 −16.9 6.5 −0.1 0.3
0.10 0.8 0 12 5 5 −27.0 2.6 −1.5 0.0
0.10 0.8 0 9 7 4 −29.7 3.8 −1.0 0.1