Table 1. Estimated type I error using the 95th quantile of the testing scores from datasets with 400 samples.

Type I error rate	n = 200	n = 400	n = 800	n = 1600
2-way
m = 10	4.8%	5.4%	5.8%	5.9%
m = 20	5.2%	5.1%	6.1%	5.5%
m = 50	5.5%	5.3%	5.8%	6.2%
3-way
m = 10	4.8%	4.1%	5.6%	6.2%
m = 20	5.3%	5.8%	5.9%	5.3%
m = 50	4.2%	4.4%	6.1%	5.4%
4-way
m = 10	4.1%	4.0%	4.9%	6.3%
m = 20	4.5%	4.8%	4.9%	4.5%
m = 50	4.1%	4.3%	4.6%	4.7%