Methods in Analyzing Missing Data in Monitoring Populations

• Activity Number: 428
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Michael Soma*+ and Doug Johnson
• Companies: U.S. Geological Survey and U.S. Geological Survey
• Keywords: missing data ; Monitoring ; ecology

Abstract

Despite paying great attention to monitoring protocols, ecologists routinely are forced to analyze collected data in which not all monitoring sites are examined on every occasion. How such missing data are handled can dramatically affect conclusions from the monitoring program. Here we evaluate several imputation methods under various common monitoring conditions. We generated data with characteristics typical of monitoring results, simulated several patterns of missingness, and examined the performance of various imputation methods. To compare how the missing data affected our estimates, we tested each scenario under each imputation method and estimated mean absolute error from 100 simulated experiments. We found that no single method worked best in all cases, but of the methods we examined, a linear model estimated from the fully observed data usually outperformed the other methods.