Archived

This content is available here strictly for research, reference, and/or recordkeeping and as such it may not be fully accessible. If you work or study at University of Kentucky and would like to request an accessible version, please use the SensusAccess Document Converter.

Publication Date

1981

Description

Conventional statistical analyses require balanced data sets. Unbalanced data can be analyzed using regression procedures, but excessive adjustments may be needed for some treatment means. The objective of this paper is to evaluate an analysis of grazing-trial data using generalized least squares. The data was obtained from a 6-year study involving 12 treatments with not all treatments being made in all years. The analyses involved estimating variance components due to years, fields, and random error and then using these estimates to obtain adjusted treatment means. Variance-component estimates for the complete data set were similar to estimates from two balanced subsets of the data. For the most part, adjusted treatment means for cow-and-calf daily gain, gain/ha, and total digestible nutrients (TDN)/ha were similar to simple treatment means. Large differences that oc­curred had rational explanations. The statistical procedure used offers considerable promise for analyzing data from unbalanced experiments involving both fixed and random effects and large random disturbances. Further experience with this analysis using both real and simulated data is needed.

Archival?

Archival

Share

COinS
 

Methodology for Determining Pasture and Animal Response Differences in Unbalanced Experiments with Intercorrelations Among Treatments

Conventional statistical analyses require balanced data sets. Unbalanced data can be analyzed using regression procedures, but excessive adjustments may be needed for some treatment means. The objective of this paper is to evaluate an analysis of grazing-trial data using generalized least squares. The data was obtained from a 6-year study involving 12 treatments with not all treatments being made in all years. The analyses involved estimating variance components due to years, fields, and random error and then using these estimates to obtain adjusted treatment means. Variance-component estimates for the complete data set were similar to estimates from two balanced subsets of the data. For the most part, adjusted treatment means for cow-and-calf daily gain, gain/ha, and total digestible nutrients (TDN)/ha were similar to simple treatment means. Large differences that oc­curred had rational explanations. The statistical procedure used offers considerable promise for analyzing data from unbalanced experiments involving both fixed and random effects and large random disturbances. Further experience with this analysis using both real and simulated data is needed.