Handling missing data in environmental monitoring programs via multiple imputation for item nonresponse
On this page:
Environmental monitoring programs often measure multiple indicator variables as part of a survey design. Commonly, data are collected for only some of the indicators but not all of them. This is called item nonresponse. One way to handle the missing data is to omit the observations when producing statistical estimates. Unfortunately, depending on the degree of missingness, this approach can yield inefficient estimates (i.e., the variance is too large). It also ignores valuable information contained in the values of the measured indicators at a site. Multiple imputation is a technique that “fills in” these missing values with plausible values based on other measured indicators and then proceeds with statistical estimation. We provide relevant background for multiple imputation and show how it can greatly reduce the uncertainty associated with estimates constructed from data having high degrees of missingness. We also discuss practical considerations for incorporating multiple imputation in environmental monitoring programs.