Recommendations for choosing an analysis method that controls Type I error for unbalanced cluster sample designs with Gaussian outcomes