Bootstrap computation of standard error for maxlogL class objects.

bootstrap_maxlogL computes standard errors of maxlogL class objects by non-parametric bootstrap.

Usage

bootstrap_maxlogL(object, R = 2000, silent = FALSE, ...)

Arguments

object: an object of maxlogL class whose standard errors are going to be computed by bootstrap.
R: numeric. It is the number of resamples performed with the dataset in bootstrap computation. Default value is 2000.
silent: logical. If TRUE, notifications of bootstrap_maxlogL are suppressed.
...: arguments passed to boot used in this routine for estimation of standard errors.

Value

A modified object of class maxlogL.

Details

The computation performed by this function may be invoked when Hessian from optim and hessian fail in maxlogL or in maxlogLreg.

However, this function can be run even if Hessian matrix calculation does not fails. In this case, standard errors in the maxlogL class object is replaced.

References

Canty A, Ripley BD (2017). boot: Bootstrap R (S-Plus) Functions.

Author

Jaime Mosquera Gutiérrez, jmosquerag@unal.edu.co

Examples

library(EstimationTools)

#--------------------------------------------------------------------------------
# First example: Comparison between standard error computation via Hessian matrix
# and standard error computation via bootstrap

N <- rbinom(n = 100, size = 10, prob = 0.3)
phat1 <- maxlogL(x = N, dist = 'dbinom', fixed = list(size = 10),
                link = list(over = "prob", fun = "logit_link"))

## Standard error computation method and results
print(phat1$outputs$StdE_Method)   # Hessian
#> [1] "Hessian from optim"
summary(phat1)
#> _______________________________________________________________
#> Optimization routine: nlminb 
#> Standard Error calculation: Hessian from optim 
#> _______________________________________________________________
#>       AIC     BIC
#>   337.521 337.521
#> _______________________________________________________________
#>      Estimate  Std. Error Z value Pr(>|z|)    
#> prob   0.28200    0.01423   19.82   <2e-16 ***
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> _______________________________________________________________
#> Note: p-values valid under asymptotic normality of estimators 
#> ---

## 'bootstrap_maxlogL' implementation
phat2 <- phat1                   # Copy the first 'maxlogL' object
bootstrap_maxlogL(phat2, R = 100)
#> 
#> ...Bootstrap computation of Standard Error. Please, wait a few minutes...
#> 
#> 
#>  --> Done <--- 

## Standard error computation method and results
print(phat2$outputs$StdE_Method)   # Bootstrap
#> [1] "Bootstrap"
summary(phat2)
#> _______________________________________________________________
#> Optimization routine: nlminb 
#> Standard Error calculation: Bootstrap 
#> _______________________________________________________________
#>       AIC     BIC
#>   337.521 337.521
#> _______________________________________________________________
#>      Estimate  Std. Error Z value Pr(>|z|)    
#> prob   0.28200    0.01354   20.82   <2e-16 ***
#> ---
#> Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
#> _______________________________________________________________
#> Note: p-values valid under asymptotic normality of estimators 
#> ---


#--------------------------------------------------------------------------------

Bootstrap computation of standard error for `maxlogL` class objects.