Functions to summarize inconvenient rjags or jagsUI coda outputs. Especially helpful when your coda summary contains timeseries variables of different lengths (e.g., when calculating posteriors for variables at daily, weekly, and seasonal timescales within the same model framework). The main function of this package is to streamline the model fitting/ output summarizing process with the least amount of thinking possible. Many functions included here are wrapper functions around functions from the postjags package developed by Michael Fell (https://github.com/fellmk/PostJAGS).
To install the package add the following code to your R script:
install.packages("devtools")
library(devtools)
First download Mike Fell's postjags package, which is a dependency:
devtools::install_github("fellmk/PostJAGS/postjags")
library(postjags)
Then download the coda4dummies package:
devtools::install_github("egreich/coda4dummies", build_vignettes = TRUE)
library(coda4dummies)
Note that you will have to load postjags with library(postjags) everytime you use coda4dummies.
# Organize the coda object as a dataframe
df_model <- dumsum(jagsobj = jagsui, type = "jagsUI")
# Connect dates to the dataframe by variable
df_model <- dateconnect(dfobj = df_model, datevect = DOY, datename = "DOY", identifier = "ID2", varlist = c("dYdX"))
df_model <- dateconnect(dfobj = df_model, datevect = YIN$TIMESTAMP, datename = "TIMESTAMP", identifier = NULL, varlist = c("WUE.pred", "T.ratio", "T.pred", "ET.pred"))
# inits to save
init_names = names(initslist[[1]]) # get the initial names from the list of initials you used to start the model
# create a saved_state object with initials for next run
# saved_state[[1]] will be the names of the initials
# saved_state[[2]] will be the list of initials
# saved_state[[3]] will be the chain number with lowest deviance
saved_state <- keepvars(codaobj = jagsui, to_keep = init_names, paramlist = params, type="jagsUI")
save(saved_state, file = initfilename) # saving the saved_state locally
load(initfilename) # loading the saved_state
# Create new initials based on the chain with the lowest deviance
new_saved_state <- lowdevrestart(saved_state = saved_state, vary_by = 10)
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent