blim: Bayesian Linear Model
In vankesteren/blim: Univariate Bayesian Linear Modelling

Description Usage Arguments Value Author(s) References See Also Examples

Fit a univariate Bayesian linear regression model.

blim(formula, data, iter = 9999, burnin = 50, inits = NULL, thin = 0
     prior_tau = "dgamma(0.001,0.001)", prior_b = "dnorm(0,10000)",
     method = "cppbr", mtsprior = F, verbose = T, dtuning = F)

## S3 method for class 'blim'
summary(blimfit, ...)

`formula`	a symbolic description of the model to be fit.
`data`	an optional data frame containing the variables in the model.
`iter`	amount of iterations for the sampler to perform and return.
`burnin`	burn-in amount of iterations ignored in the trace.
`inits`	initial values for variance and betas in a `"list"` format.
`thin`	amount of thinning to be performed
`prior_tau`	prior for the residual variance in the linear model, of the form dgamma(shape, scale). Only dgamma allowed!
`prior_b`	(vector of) priors for beta coefficients. With `method = "rmhs"`, nonconjugate priors are allowed. Any density function works, as long as it has a d and an r method.
`method`	cppbr = Bayes Regression in C++, rbr = Bayes Regression in R, rgs = Bayesian Regression under Gibbs Sampling, rmhs = Bayesian Regression under Gibbs Sampling with Metropolis-Hastings step for Nonconjugate Priors
`mtsprior`	Boolean value indicating whether a Minimum Training Sample prior should be calculated. For use in conjuction with ANOVA & Bayes Factors for informative hypotheses
`verbose`	print all info in the console or not (in development!)
`dtuning`	dynamic tuning of the random walk Metropolis-Hastings step

An object of class blimfit, including all necessary information such as summary, trace and original data.

Erik-Jan van Kesteren

Beland, S., Klugkist, I., Raiche, G. & Magis, D. (2012). A short introduction into Bayesian evaluation of informative hypotheses as an alternative to exploratory comparisons of multiple group means. Tutorials in Quantitative Methods for Psychology 8(2), p. 122-126.

Raftery, A. E. (1996). Hypothesis testing and model selection. Markov chain Monte Carlo in practice, 163-188.

cplots, aplots, BF, PPC, ICBF, DIC

# Most simple example with full functionality
fit1 <- blim(dist~speed, data=cars)

summary(fit1)

# More complex example with nonconjugate priors
fit2 <- blim(formula = dist~speed, data = cars, iter = 9999, burnin = 100,
             inits = list(var = 200, b0 = -15, b1 = 3), thin = 0,
             prior_tau = "dgamma(0.001,0.001)",
             prior_b = c("dcauchy(0,2000)", "dunif(3,4.5)"),
             method = "rmhs", dtuning = 1000)

summary(fit2)

# check convergence and autocorrelation
cplots(fit1)
aplots(fit1)

# calculate Bayes Factor of model 2 versus model 1
BF(fit2, fit1)

# perform posterior predictive check for normality of residuals
PPC(fit2)

# evaluate informative hypothesis in model 1
ICBF(fit1, model = "par[1] < par[2]", complement = TRUE)