Statistical tools: experimental design and monitoring

Outlier Identification

Robust methods for outlier detection are often required in plant phenotyping facilities. Based on interactions with EPPN2020 platform managers gidelines and tools for automating the steps of data analysis have been developed and are integrated in the R package statgenHTP providing a suite of functions to:

• detect outliers at the level of individual plant observations and at the level of time series for a trait at a plant
• separate the genetic effects from the spatial effects at each time point
• estimate parameters from a time course modelled with splines

Experimental Design Generator

Phenotyping platforms require a reconsideration of classical experimental design and analysis techniques currently used for field experiments. It is not widely recognised, spatial and temporal heterogeneities in controlled conditions are large, often as large as in the field. It is essential that users choose appropriate experimental designs, models and analysis methods. Thus, platform experiment needs to follow standard principles for experimental design, incl.:

• Replication: to allow quantifying the experimental variation between experimental units and increase the precision of the estimated effect
• Randomization: to avoid confounding of treatment difference and (unknown) other differences between (groups of) units.
• Restriction: to reduce the experimental error by grouping experimental units into blocks of similar experimental conditions (homogenious)