Skip to main content
ARS Home » Research » Publications at this Location » Publication #164162
Title: ESTIMATION OF DURATION INDICES FOR TENSIOMETER REPEATED MEASURES

Author
item Meek, David
item Singer, Jeremy

Submitted to: Agronomy Journal
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/1/2004
Publication Date: 11/17/2004
Citation: Meek, D.W., Singer, J.W. 2004. Estimation of duration indices for tensiometer repeated measures. Agronomy Journal. 96(6):1787-1790.

Interpretive Summary: Observations collected in scientific studies are often repeated over time resulting in a series of observations per subject. The series of observations can provide more and different information than a single observation in time does about a subject's response to a given treatment. Not surprisingly, the comparison of treatments between two or more subjects having a series of observations is more complicated than one that has a single datum per subject. The persistence of the observed response was our interest. The ususal analysis is both complicated and laborious. So we compared it to a simpler approach for calculating the persistence of the observations repeated over time; results were the same. The simpler technique will help interested researchers save time in their analyses of repeated observations.

Technical Abstract: Regression analysis of repeated measures data from plots in large experiments is both labor intensive and time consuming. When a duration estimate is of interest, numerical integrations are a possible alternative. This note evaluates a trapezoidal rule wetness duration index derived from tensiometer repeated measures associated with a recent soybean [Glycine max (L.) Merr.] study.Tensiometer repeated measures series were randomly selected from four of the treatments with their replications for a total of 15 plots (of 420 total) associated with a 2 year experiment. Regression based duration estimates were developed and compared to those derived using an unequally spaced trapezoidal rule. Equivalence between the two estimators was assessed based on multiple statistical analyses. The estimators were virtually identical by all criteria. The numerical method is faster and easier to employ. Furthermore, the concept can be applied to other repeated measurements and the results may be useful as covariates in other analyses.