The Complexities of a Multivariate Power Analysis: Possible Finite Solutions to an Infinite Problem

• Activity Number: 161
• Type: Topic Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Kathryn Mary Irvine*+ and Daniel Sarr and Eric Dinger
• Companies: Montana State University and Klamath Network - National Park Service and Klamath Network - National Park Service
• Keywords: Progressive Trend ; Mantel Test ; PerMANOVA ; species composition ; Monitoring

Abstract

Progressive trend is defined as increasing/decreasing dissimilarity in species composition over time. This multivariate alternative hypothesis for trend presents challenges for developing a power analysis. We explore an extension to a multi-species occupancy model that allows for dissimilar trends amongst species groups for simulating data under the alternative. We determine if such a model produces simulated data that mimics theoretical ecological scenarios of interest. We investigate the influence of number of species, magnitude of trend in univariate species, number of sites, and number of years on the power to detect progressive trend. We explore both PerMANOVA and the Mantel test for detecting progressive trend. We discuss our results in the context of developing robust sampling designs for monitoring multi-species assemblages within the national parks of the Klamath Network.