Upper Midwest Environmental Sciences Center
Long Term Resource Monitoring Program
LTRMP Statistics Estimating means and temporal trends using LTRMP data: details with examplesEstimating temporal trends
The LTRMP provides data collected with consistent methods over many years. Challenges associated with estimating trends from data like ours are addressed at Estimating Trends in LTRMP Data.
The subsequent pages provide code for estimating temporal trends that requires minimal modeling assumptions. The method proposed for estimating trends from continuous data (water quality and fish length) uses generalized least squares, which does not make a parametric distributional assumption for observed data. By contrast, the methods proposed for count and categorical data do make distributional assumptions (Poisson and logistic, respectively--conditional on annual changes in means). However, none of the proposed methods make distributional assumptions about the distribution of annual means (as would standard parametric models of clustered data). For these reasons, we view the methods proposed here as conservative and, hence, reasonable for an ecological monitoring program.
The following pages address the estimation of temporal trends within and among strata.
Contact: Questions or comments may be directed to Brian Gray, LTRMP statistician, Upper Midwest Environmental Sciences Center, La Crosse, Wisconsin, at brgray@usgs.gov.
URL: http://www.umesc.usgs.gov/stats/estimating_temporal_trends.html
