Description
Critical findings on design, statistical analysis, and interpretation of the results will be addressed based on comparative ensiling trials. For this aim, a lab-scale ensiling trial on biostatistical issues was conducted in 2021. Grass material from a permanent mowing pasture was taken from (i) 10 sampling points, (ii) one sampling point, (iii) a mixture of 10 sampling points. For each sub-trial (based on the sampling design), 3 levels of the fixed treatment factor silage additive were tested with 10 replicates (without additive, chemical silage additive, biological silage additive). The analysis was performed within a linear mixed effects model (LMM) as randomized complete block design (RCBD), accounting for systematic effects of field sampling points (i) and/or time processing (i, ii, iii). In sub- trial (i), variability in trait values was highest and more influenced by treatments (variance heterogeneity), and block effects were most pronounced. In contrast, the block effect was less pronounced in (ii) and (iii), and we could not find a time gradient in the silage trait values. Depending on the nature of the silage trait (distribution, treatment variances), a suitable analysis procedure has to be chosen. The frequently used low number of replications is probably not sufficient.
Included in
Agricultural Science Commons, Agronomy and Crop Sciences Commons, Plant Biology Commons, Plant Pathology Commons, Soil Science Commons, Weed Science Commons
Considerations on Sampling and Statistical Analysis in Grassland Ensiling Trials
Critical findings on design, statistical analysis, and interpretation of the results will be addressed based on comparative ensiling trials. For this aim, a lab-scale ensiling trial on biostatistical issues was conducted in 2021. Grass material from a permanent mowing pasture was taken from (i) 10 sampling points, (ii) one sampling point, (iii) a mixture of 10 sampling points. For each sub-trial (based on the sampling design), 3 levels of the fixed treatment factor silage additive were tested with 10 replicates (without additive, chemical silage additive, biological silage additive). The analysis was performed within a linear mixed effects model (LMM) as randomized complete block design (RCBD), accounting for systematic effects of field sampling points (i) and/or time processing (i, ii, iii). In sub- trial (i), variability in trait values was highest and more influenced by treatments (variance heterogeneity), and block effects were most pronounced. In contrast, the block effect was less pronounced in (ii) and (iii), and we could not find a time gradient in the silage trait values. Depending on the nature of the silage trait (distribution, treatment variances), a suitable analysis procedure has to be chosen. The frequently used low number of replications is probably not sufficient.
