This consulting project was part of my graduate consulting curriculum, where students provide statistical consulting for the statistics department.
The client (Patrick Michelsen) was an Environmental Management M.S. student working on their thesis project. Patrick needed guidance on what statistical method to use for his randomized complete block design study with 6 response values. His main issue was he has many missing values and values under the limit of detection (LOD), aka left censored.
My group presented two approaches:
Using Bayesian Statistics with the
brmslibrary, which allowed us to use a mixed model with left censored dataTreating the LOD values as 0 and using a Zero-Inflated model to separately model the LOD and actual values
For the two response variables without any LOD values, we presented generalized linear mixed models.