Bootstrap simulation of uncertainty for non-censored data

Uses parametric or nonparametric bootstrap resampling in order to simulate uncertainty in the parameters of the distribution fitted to non-censored data.

Usage

bootdist(f, bootmethod = "param", niter = 1001, silent = TRUE, 
      parallel = c("no", "snow", "multicore"), ncpus)
# S3 method for class 'bootdist'
print(x, ...)
# S3 method for class 'bootdist'
plot(x, main = "Bootstrapped values of parameters", enhance = FALSE, 
    trueval = NULL, rampcol = NULL, nbgrid = 100, nbcol = 100, ...)
# S3 method for class 'bootdist'
summary(object, ...)
# S3 method for class 'bootdist'
density(..., bw = nrd0, adjust = 1, kernel = "gaussian")
# S3 method for class 'density.bootdist'
plot(x, mar=c(4,4,2,1), lty=NULL, col=NULL, lwd=NULL, ...)
# S3 method for class 'density.bootdist'
print(x, ...)

Arguments

f: An object of class "fitdist", output of the fitdist function.
bootmethod: A character string coding for the type of resampling : "param" for a parametric resampling and "nonparam" for a nonparametric resampling of data.
niter: The number of samples drawn by bootstrap.
silent: A logical to remove or show warnings and errors when bootstraping.
parallel: The type of parallel operation to be used, "snow" or "multicore" (the second one not being available on Windows), or "no" if no parallel operation.
ncpus: Number of processes to be used in parallel operation : typically one would fix it to the number of available CPUs.
x: An object of class "bootdist" or "density.bootdist".
object: An object of class "bootdist".
main: an overall title for the plot: see title, default to "Bootstrapped values of parameters".
enhance: a logical to get an enhanced plot.
trueval: when relevant, a numeric vector with the true value of parameters (for backfitting purposes).
rampcol: colors to interpolate; must be a valid argument to colorRampPalette().
nbgrid: Number of grid points in each direction. Can be scalar or a length-2 integer vector.
nbcol: An integer argument, the required number of colors
...: Further arguments to be passed to generic methods or "bootdist" objects for density.
bw, adjust, kernel: resp. the smoothing bandwidth, the scaling factor, the kernel used, see density.
mar: A numerical vector of the form c(bottom, left, top, right), see par.
lty, col, lwd: resp. the line type, the color, the line width, see par.

Details

Samples are drawn by parametric bootstrap (resampling from the distribution fitted by fitdist) or nonparametric bootstrap (resampling with replacement from the data set). On each bootstrap sample the function mledist (or mmedist, qmedist, mgedist according to the component f$method of the object of class "fitdist") is used to estimate bootstrapped values of parameters. When that function fails to converge, NA values are returned. Medians and 2.5 and 97.5 percentiles are computed by removing NA values. The medians and the 95 percent confidence intervals of parameters (2.5 and 97.5 percentiles) are printed in the summary. If inferior to the whole number of iterations, the number of iterations for which the function converges is also printed in the summary.

By default (when enhance=FALSE), the plot of an object of class "bootdist" consists in a scatterplot or a matrix of scatterplots of the bootstrapped values of parameters. It uses the function stripchart when the fitted distribution is characterized by only one parameter, the function plot when there are two paramters and the function pairs in other cases. In these last cases, it provides a representation of the joint uncertainty distribution of the fitted parameters.

When enhance=TRUE, a personalized plot version of pairs is used where upper graphs are scatterplots and lower graphs are heatmap image using image based on a kernel based estimator for the 2D density function (using kde2d from MASS package). Arguments rampcol, nbgrid, nbcol can be used to customize the plots. Defautls values are rampcol=c("green", "yellow", "orange", "red"), nbcol=100 (see colorRampPalette()), nbgrid=100 (see kde2d). In addition, when fitting parameters on simulated datasets for backtesting purposes, an additional argument trueval can be used to plot a cross at the true value.

It is possible to accelerate the bootstrap using parallelization. We recommend you to use parallel = "multicore", or parallel = "snow" if you work on Windows, and to fix ncpus to the number of available processors.

density computes the empirical density of bootdist objects using the density function (with Gaussian kernel by default). It returns an object of class density.bootdist for which print and plot methods are provided.

Value

bootdist returns an object of class "bootdist", a list with 6 components,

estim: a data frame containing the bootstrapped values of parameters.
converg: a vector containing the codes for convergence obtained if an iterative method is used to estimate parameters on each bootstraped data set (and 0 if a closed formula is used).
method: A character string coding for the type of resampling : "param" for a parametric resampling and "nonparam" for a nonparametric resampling.
nbboot: The number of samples drawn by bootstrap.
CI: bootstrap medians and 95 percent confidence percentile intervals of parameters.
fitpart: The object of class "fitdist" on which the bootstrap procedure was applied.

Generic functions:

print: The print of a "bootdist" object shows the bootstrap parameter estimates. If inferior to the whole number of bootstrap iterations, the number of iterations for which the estimation converges is also printed.
summary: The summary provides the median and 2.5 and 97.5 percentiles of each parameter. If inferior to the whole number of bootstrap iterations, the number of iterations for which the estimation converges is also printed in the summary.
plot: The plot shows the bootstrap estimates with stripchart function for univariate parameters and plot function for multivariate parameters.
density: The density computes empirical densities and return an object of class density.bootdist.

References

Cullen AC and Frey HC (1999), Probabilistic techniques in exposure assessment. Plenum Press, USA, pp. 181-241.

Delignette-Muller ML and Dutang C (2015), fitdistrplus: An R Package for Fitting Distributions. Journal of Statistical Software, 64(4), 1-34, doi:10.18637/jss.v064.i04 .

Author

Marie-Laure Delignette-Muller and Christophe Dutang.

Examples

# We choose a low number of bootstrap replicates in order to satisfy CRAN running times
# constraint.
# For practical applications, we recommend to use at least niter=501 or niter=1001.


# (1) Fit of a gamma distribution to serving size data
# using default method (maximum likelihood estimation)
# followed by parametric bootstrap
#
data(groundbeef)
x1 <- groundbeef$serving
f1 <- fitdist(x1, "gamma")
b1 <- bootdist(f1, niter=51)
print(b1)
#> Parameter values obtained with parametric bootstrap 
#>      shape       rate
#> 1 4.081548 0.05486224
#> 2 4.189530 0.05717646
#> 3 4.186561 0.05827691
#> 4 4.153367 0.06010613
#> 5 5.011150 0.07210224
#> 6 4.332796 0.05647597
plot(b1)

plot(b1, enhance=TRUE)

summary(b1)
#> Parametric bootstrap medians and 95% percentile CI 
#>           Median       2.5%      97.5%
#> shape 4.11174269 3.33866627 5.05787399
#> rate  0.05532224 0.04714105 0.07039183
quantile(b1)
#> (original) estimated quantiles for each specified probability (non-censored data)
#>             p=0.1    p=0.2    p=0.3    p=0.4    p=0.5    p=0.6    p=0.7
#> estimate 32.16733 42.32692 50.91831 59.15298 67.62801 76.88308 87.67764
#>             p=0.8    p=0.9
#> estimate 101.5208 122.9543
#> Median of bootstrap estimates
#>             p=0.1    p=0.2    p=0.3    p=0.4    p=0.5    p=0.6    p=0.7
#> estimate 32.65191 42.77486 51.16946 59.25783 67.54032 76.42359 87.07212
#>             p=0.8    p=0.9
#> estimate 100.8478 121.5045
#> 
#> two-sided 95 % CI of each quantile
#>           p=0.1    p=0.2    p=0.3    p=0.4    p=0.5    p=0.6    p=0.7     p=0.8
#> 2.5 %  27.78104 37.70176 45.79023 53.77101 62.30589 71.51079 81.58248  93.57508
#> 97.5 % 35.20921 45.25381 53.86981 62.27518 71.23506 81.28761 93.12187 107.81119
#>           p=0.9
#> 2.5 %  112.6206
#> 97.5 % 130.6195
CIcdfplot(b1, CI.output = "quantile")

density(b1)
#> 
#> Bootstrap values for: gamma for 1 object(s) with 51 bootstrap values (original sample size 254).
plot(density(b1))


# (2) non parametric bootstrap on the same fit
#
b1b <- bootdist(f1, bootmethod="nonparam", niter=51) 
summary(b1b)
#> Nonparametric bootstrap medians and 95% percentile CI 
#>           Median       2.5%      97.5%
#> shape 4.07620083 3.45714119 4.70574875
#> rate  0.05554833 0.04814463 0.06355013
quantile(b1b)
#> (original) estimated quantiles for each specified probability (non-censored data)
#>             p=0.1    p=0.2    p=0.3    p=0.4    p=0.5    p=0.6    p=0.7
#> estimate 32.16733 42.32692 50.91831 59.15298 67.62801 76.88308 87.67764
#>             p=0.8    p=0.9
#> estimate 101.5208 122.9543
#> Median of bootstrap estimates
#>             p=0.1    p=0.2    p=0.3    p=0.4    p=0.5    p=0.6    p=0.7
#> estimate 32.15396 42.37584 50.96144 58.92657 67.42331 76.79944 87.14609
#>             p=0.8   p=0.9
#> estimate 100.2863 121.457
#> 
#> two-sided 95 % CI of each quantile
#>           p=0.1    p=0.2    p=0.3    p=0.4    p=0.5    p=0.6    p=0.7     p=0.8
#> 2.5 %  28.98195 38.64795 47.20053 55.45966 63.67865 72.26101 82.41641  95.36231
#> 97.5 % 36.61241 46.92714 55.51564 63.66138 71.97465 80.97711 91.55170 105.36847
#>           p=0.9
#> 2.5 %  115.1418
#> 97.5 % 127.5526


# (3) Fit of a normal distribution on acute toxicity values of endosulfan in log10 for
# nonarthropod invertebrates, using maximum likelihood estimation
# to estimate what is called a species sensitivity distribution 
# (SSD) in ecotoxicology, followed by estimation of the 5 percent quantile value of 
# the fitted distribution, what is called the 5 percent hazardous concentration (HC5)
# in ecotoxicology, with its two-sided 95 percent confidence interval calculated by 
# parametric bootstrap
#
data(endosulfan)
ATV <- subset(endosulfan, group == "NonArthroInvert")$ATV
log10ATV <- log10(subset(endosulfan, group == "NonArthroInvert")$ATV)
fln <- fitdist(log10ATV, "norm")
bln <- bootdist(fln, bootmethod = "param", niter=51)
quantile(bln, probs = c(0.05, 0.1, 0.2))
#> (original) estimated quantiles for each specified probability (non-censored data)
#>            p=0.05    p=0.1  p=0.2
#> estimate 1.744227 2.080093 2.4868
#> Median of bootstrap estimates
#>            p=0.05   p=0.1    p=0.2
#> estimate 1.927721 2.24123 2.570288
#> 
#> two-sided 95 % CI of each quantile
#>           p=0.05    p=0.1    p=0.2
#> 2.5 %  0.9768477 1.397718 1.937638
#> 97.5 % 2.4716633 2.703313 2.983822

# (4) comparison of sequential and parallel versions of bootstrap
# to be tried with a greater number of iterations (1001 or more)
#
# \donttest{
niter <- 1001
data(groundbeef)
x1 <- groundbeef$serving
f1 <- fitdist(x1, "gamma")

# sequential version
ptm <- proc.time()
summary(bootdist(f1, niter = niter))
#> Parametric bootstrap medians and 95% percentile CI 
#>           Median       2.5%      97.5%
#> shape 4.02419491 3.47031893 4.73319322
#> rate  0.05457466 0.04623237 0.06478125
proc.time() - ptm
#>    user  system elapsed 
#>   4.167   0.002   4.171 

# parallel version using snow
require("parallel")
#> Loading required package: parallel
ptm <- proc.time()
summary(bootdist(f1, niter = niter, parallel = "snow", ncpus = 2))
#> Parametric bootstrap medians and 95% percentile CI 
#>           Median       2.5%      97.5%
#> shape 4.02295676 3.44821335 4.75696688
#> rate  0.05449994 0.04620417 0.06557893
proc.time() - ptm
#>    user  system elapsed 
#>   0.039   0.004   3.918 

# parallel version using multicore (not available on Windows)
ptm <- proc.time()
summary(bootdist(f1, niter = niter, parallel = "multicore", ncpus = 2))
#> Parametric bootstrap medians and 95% percentile CI 
#>           Median       2.5%      97.5%
#> shape 4.04594150 3.47586081 4.74328466
#> rate  0.05490323 0.04682605 0.06512702
proc.time() - ptm
#>    user  system elapsed 
#>   4.198   0.298   2.319 
# }