Let’s say that you’re fitting a cumbersome model so time is not to waste over a PC staring at the screen half anxious-half bored…
Then, you can always leave and go on with meetings and all your daily routine and have R notify you the results! How?
We will illustrate the situation above using some Bayesian Model Averaging code adapted by Martin Feldkircher & Stefan Zeugner. You should download the code and source everything in R except for the example in the end (after the definition of the functions!).
#The code to get a model fls.data=read.table(url("http://feldkircher.gzpace.net/links/fls_data_adj.txt")) data.M=as.matrix(fls.data) K=ncol(dataM)-1 # nr. of regressors # this setting corresponds to a uniform prior on the model space (prior.msize=K/2 and theta="fix") # and the ric specification since K^2> N (with N the nr. of observations) as suggested by fls model.ric=fls(X.data=data.M,burn=60000,iter=700000,g=(1/K^2),nmodel=100,theta="fix",prior.msize=K/2,logfile=T,mcmc="bd",start.value=rep(0,K),beta.save=T)
This is gonna take s o m e time (really!), so you could let R working and go out for a cup of coffee (typical of Greek people!). Add the following at the end of the above code.
library(twitteR) sess <- initSession('myUser', 'myPass') # Set your user account info ns <- updateStatus('A model waits for you @ home ;)', sess)
Would you really care enough to check whether the fit is done when outside?tags: averaging - bayesian - code - model - R - twitter