Skip to content

Latest commit

 

History

History
34 lines (28 loc) · 1.48 KB

README.md

File metadata and controls

34 lines (28 loc) · 1.48 KB
Raw

Hierarchical Functional Models

Hierarchical Functional Models, unified sampler with three types of priors

  • Matrix Normal (Naive Bayesian Conjugate Prior) (hfm.NB)
  • Strongly Separable Latent Factor Prior (hfm.LF with two inputs)
  • Non-Separable Latent Factor Prior (hfm.LF with one input)

Installation:

# For Mac Users
devtools::install_github("Qian-Li/HFM")
# For Windows and Linux Users, please try install the binary version

Updates (March 6th, 2018):

  • Unified implementation across all priors;
  • Modified separable simulation: simulated B-spline coefficients directly instead of GP observations;
  • Second-moment assessment (Random effects)
  • Simulation 1: Random missing observations on varying bs.df (num of spline coefs); Sep and Non-sep 500
  • Simulation 2: Random missing observations on varying spatial dependency (rho = .4/.6/.8); Sep and Non-sep 500
  • Simulation 3: Varying Sample size and SNR; (100);
  • Simulation 4: Robustness of noLF when true noLF unknown; (100);

Updates (March 29th, 2018):

  • Data Analysis: Alpha and Gamma band regressed on Groups;
  • Data Analysis: Alpha and Gamma band regressed on Groups and Age;
  • Data Analysis: Alpha and Gamma band regressed on Age and VDQ;

Updates (May 3rd, 2018):

  • 1.0 Published: all issues fixed and sampling progrssion bar added :-)
  • Coming soon: A vignette and short working example of how to use the package

Debug:

  • Crash occasionally on pinv() or chol();