bootglg function

Bootstrap inference for a generalized log-gamma regression

bootglg is used to generate parametric bootstrap inference, such as, estimated standard errors and approximate confidence intervals for a generalized log-gamma regression.

bootglg(formula, data, B = 500, alpha = 0.05, type = "normal", plt_den = FALSE)

Arguments

• formula: a symbolic description of the systematic component of the model to be fitted.
• data: data.frame, contains the variables in the formula object.
• B: integer, represents the number of bootstrap replications. Default value is 500.
• alpha: numeric, represents a confidence level for the bootstrap intervals. Default value is 0.05.
• type: character, indicates the type of bootstrap confidence interval for the estimated parameters. The options are: 'normal', 't_student' or 'bootstrap_t'. These intervals used the bootstrap estimated standard error of the ML estimates of the parameters. Other kind of bootstrap intervals are the percentile-type intervals. We offer the option 'BCa'. It is a bias-corrected and accelerated percentile interval. The default value for the 'type' argument is 'normal'.
• plt_den: boolean value, to request a density-type plot of the bootstrap estimates. Default value is FALSE.

Returns

ml_estimates is a vector of maximum likelihood estimates associated with the coefficients of linear structure, scale, and shape parameters.

boot_mean_estimates is a vector of mean of the bootstrap estimates associated with the coefficients of linear structure, scale, and shape parameters.

boot_bias_estimates is a vector of bootstrap estimate of bias associated with the coefficients of linear structure, scale, and shape estimators.

boot_sd_estimates is a vector of bootstrap standard errors of the estimates associated with the coefficients of linear structure, scale, and shape estimators.

type indicates the type of confidence intervals.

intervals array of the confidence intervals of the coefficients of linear structure, scale, and shape.

Examples

##################################################################################################
set.seed(1)
n <- 300
x1 <- runif(n, 0, 1)
t_beta  <- 1.2
t_sigma <- 0.5
t_lambda <- 0.7
error <- rglg(n, 0, t_sigma, t_lambda)
y1 <- t_beta*x1 + error
data <- data.frame(y1, x1)
# The following examples are based on 50 bootstrap replications.
# A 90% bootstrap confidence interval with the method 'normal'.
bootglg(y1 ~ x1 - 1, data = data, type='normal', B = 50, alpha = 0.1)
# A 95% bootstrap confidence interval with the method 't_student'.
bootglg(y1 ~ x1 - 1, data = data, type='t_student', B = 50)
# A 95% bootstrap confidence interval with the method 'bootstrap_t'.
bootglg(y1 ~ x1 - 1, data = data, type='bootstrap_t', B = 50)
# A 98% bootstrap confidence interval with the method 'BCa'.
# bootglg(y1 ~ x1 - 1, data = data, type='BCa', B = 50, alpha = 0.02)
#################################################################################################

References

Cardozo C. A., Paula G. and Vanegas L. sglg: An R package to fit semi-parametric generalized log-gamma regression models. In preparation.

Efron B and Tibshirani R (1993). An introduction to the Bootstrap. Chapman & Hall, Inc.

Author(s)

Carlos Alberto Cardozo Delgado cardozorpackages@gmail.com