Skip to main content
. Author manuscript; available in PMC: 2018 Jun 5.
Published in final edited form as: Ophthalmic Epidemiol. 2017 Sep 11;25(1):45–54. doi: 10.1080/09286586.2017.1339809

Table 1.

Type I error rates of the statistical tests in the simulation study at various scenarios for comparing the means between two groups when two eyes of a subject are in two different comparison groups.

ρ No. of subjects Type I error ratea

Two sample t-test Paired t-test GEE Wald test GEE Score test LMM
0 5 0.056 0.050 0.15 0.028 0.050
10 0.048 0.052 0.10 0.043 0.052
20 0.051 0.054 0.080 0.051 0.054
30 0.051 0.053 0.064 0.052 0.053
50 0.046 0.050 0.056 0.050 0.050
0.25 5 0.035 0.056 0.15 0.026 0.056
10 0.026 0.052 0.10 0.040 0.052
20 0.026 0.054 0.077 0.050 0.054
30 0.025 0.053 0.067 0.051 0.053
50 0.023 0.050 0.054 0.049 0.050
0.5 5 0.016 0.054 0.15 0.023 0.054
10 0.014 0.051 0.10 0.042 0.051
20 0.007 0.050 0.074 0.047 0.050
30 0.007 0.049 0.066 0.047 0.049
50 0.007 0.048 0.060 0.047 0.048
0.75 5 0.003 0.055 0.16 0.025 0.055
10 0.0005 0.050 0.10 0.041 0.050
20 0.0005 0.054 0.077 0.051 0.054
30 0 0.052 0.066 0.050 0.052
50 0 0.051 0.058 0.050 0.051
a

The test with type I error rate closest to 0.05 has the best statistical performance.

ρ, Inter-eye correlation; GEE, Generalized estimating equations; LMM, Linear mixed effect model.