Review literature for bayesian models of log normal outcome with 0s

UrbanStudy / HB_GLMM

Hierarchical Bayes for Generalized Linear Mixed Effect Model

https://UrbanStudy.github.io/HB_GLMM/README.html

0 stars 0 forks source link

Review literature for bayesian models of log normal outcome with 0s #1

Open lmwang9527 opened 4 years ago

lmwang9527 commented 4 years ago

one-part model
- log_normal model that handles 0, does shifted_lognormal work?
- zero-inflated-negative-binomial or hurdle model
two-part models

qushen26 commented 4 years ago

Family 'lognormal' requires the response greater than 0 "Family exgaussian ('exponentially modified Gaussian') and shifted_lognormal are especially suited to model reaction times." shift value is needed. Family 'zero_inflated_binomial' requires integer responses hurdle_lognormal ...............works, but very long. Reports not converge.

qushen26 commented 4 years ago

Truncated issues. Rectified Normal is a mixture of constant (0) and Truncated Normal. Truncated Normal is a bouned Normal. Rethinking if negative values are unobserved part of a normal, or meaningless.