Online Program

Return to main conference page
Saturday, May 19
Computational Statistics
Bioinformatics
Sat, May 19, 10:30 AM - 12:00 PM
Regency Ballroom A
 

Type-I Error Rate of a One-Way ANOVA in the Case of a Large Number of Factors With Small Replications (304505)

Richard Opoku-Nsiah, Liberty Mutual Insurance 
*Sharad Silwal, Jefferson College of Health Sciences 
Haiyan Wang, Kansas State University 

Keywords: Edgeworth expansion, high-dimensional ANOVA, convergence rate, type-I error accuracy

With technological conveniences, large amounts of data with limited sample sizes have become common in many areas of scientific research, for instance, agricultural screening and micro-array experiments. Arkitas and Papadatos (2004) developed a one-way ANOVA test in the case of a large number of factors (a) and small replications (n_i) with heteroscedastic variances (s_i^2). We show that their type-I error accuracy is O(a^(-1/2)). By using a higher-order approximation of the distribution of the test statistic, we improve this accuracy up to O(a^(-1)). Simulation studies are included to evaluate the performance of our results.