search

NIVANorge / Milkys2_pc

R scripts and Excel templates for (1) fetching data to use in Milkys2 on Jupyterhub, and post-processing files created in Milkys2 on Jupyterhub
1 stars 2 forks source link

Milkys app (866 Tukey Kramer-test) - handle data < LOQ better #2

Open DagHjermann opened 2 years ago

DagHjermann commented 2 years ago

Either use random numbers between 0.5 and 1 LOQ, or Bayesian methods

DagHjermann commented 2 years ago

Needs to

Both of which really points towards Bayesian methods (MCMC through JAGS or Stan)
See these efforts:

  1. https://github.com/NIVANorge/espial/blob/master/91_Test_analysis_leftcensored.md (the first try)
  2. The package C:/Data/R_packages/Leftcensored (not on Github!)
  3. https://github.com/lenz99-/lme4cens (also forked by myself)
  4. Used in this script in MAR001_2021

For non-linear trends, see this repo (forked) and this blog post