|
Abstract:
|
Recent studies have compared different approaches to analysis of data with large number of measures (p) relative to n, the number of subjects (e.g., Myers, et al 2008). This type of data is found in many fields including public health, biology and anthropology. Traditional multivariate analysis methods such as MANOVA are not optimal for data with high p/n ratios so appropriate analysis of these data presents a challenge. Alternative analysis methods for testing equality of group centroids were explored using simulations. The robustness of these methods with respect to Type I error was assessed for various p to n ratios, overall sample sizes, and correlations between measures.
|
