Workshop: BLUPs – a case of using random effects as a variable of interest
Description:
This is an example activity part of the workshop on R coding and statistics that we held in the Department of Agroecology during the first semester of 2024. It serves as a prototype of the teaching methodology of a PhD curriculum in Applied Biostatistics, which is the result of the SoTL research of the AU pedagogical project.
The general methodology is based on a case study requested by one or more of the participants. We first take time to conceptualise the problem so that we can formulate a specific research question that can be translated into a specific statistical model that can answer it. We work with the data, carry out the analysis and finally discuss the implications of the results.
Intended Learning Outcomes:
- Recall on Linear Mixed Model framework for analysing experimental data
- Define the BLUPs and the BLUE concepts
- Explore utility random effect point estimates to describe the distribution of individual effects within a sampled population
- Illustrate the utility of BLUPs estimation in breeding selection
- Estimate Linear Mixed Models in R
| Resources | Tasks | Supports | |||
|---|---|---|---|---|---|
|
Previous workshops or out-of class activity |
|||||
|
Board, poster paper and markers |
Identifying and selecting the question to be answered from the student data ↓ |
← |
Student case, peer feedback from colleagues and tutors |
||
|
PC: Excel software and R software through R Studio interface |
Preparation of data structure for statistical analysis ↓ |
← |
Peer feedback from colleagues and tutors |
||
|
Workshop in-class activity |
|||||
|
PC, projector and presentation with slides and references |
Short lecture on the implementation of linear mixed models with an emphasis on the concepts of BLUEs and BLUPs. ↓ |
← |
Tutors |
||
|
Board, poster paper and markers |
Contextualise the research problem and define the specific research question to be answered by a linear mixed model. ↓ |
← |
Student case, student, peers |
||
|
Personal computer, projector and R software via R Studio interface |
Running the model, extracting BLUPs and describing them in R ↓ |
← |
Student and tutors |
||
|
Board, PC, projector and R software via R Studio interface |
Interpreting and discussing the results and their implications. ↓ |
← |
Peers, student and tutors |
||
|
Documentation out-of-class activity |
|||||
|
PC, R software via R Studio interface and github repository. |
Adapting the inclass example to a sherable toy dataset. Commenting the code and uploading it to a public repository. |
← |
Tutors |
||
Additional information
Context:
This activity in the Data and Statistical Analysis workshops respects the general structure of the rest of the session, but focuses on a specific topic. By the time this activity takes place we will have had six sessions, starting with the basics of R coding and summary statistics, building the concept of linear models and introducing the concept of random effects for at least two sessions.
We are two tutors supporting the workshop in the interaction of co-teaching is highly appreciated. The resources are the tutors' and participants' PCs, a shared classroom, preferably with large tables, and other traditional classroom tools such as whiteboard, flipchart and paper. Software resources are particularly important in this context. We work with open R software via R Studio implementations and build a library of code and examples on a publicly accessible Github repository.
Description of the specific activity:
The student presenting has to describe the experimental design that gives rise to the database, then should present the database identifying the unit of observation, the response variable and the explanatory variables that would be part of the linear model to be run. The teacher asks questions to guide the students' discussion on how to build a model to answer the specific research question. The students, sharing the computer screen, run the model according to the instructions and make an initial interpretation of the results with their peers. Then a second round of interpretation takes place with contributions from the tutors. Finally, the tutors propose recommendations on how to report the results.