How to select the distribution parameter through fitdistrplus package?

user_7936836 0
Viewed 141
library(fitdistrplus)
data(groundbeef)
serving <- groundbeef$serving
fitg <- fitdist(serving, "gamma")
res<-bootdist(fitg)#further accuracy estimate
#which parameter should be selected?
> summary(res)
Parametric bootstrap medians and 95% percentile CI 
Median       2.5%      97.5%
shape 4.04523398 3.41773958 4.80480936
rate  0.05495878 0.04632007 0.06587446
> res$fitpart$estimate
shape       rate 
4.00825257 0.05441911

bootdist() always used to get more accurate parameter after fitdis().

As above example,summary(res) provide several parameters,res$fitpart$estimate provide one parameter,which one should be selected?

fitdistrplus statistics r
2 Answers

From help(bootdist):

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

Thus, the purpose of bootdist is not to determine the parameters, but rather estimate the uncertainty in the fit of the parameters.

Therefore, in general, you should use:

fitg$estimate
     shape       rate 
4.00825257 0.05441911
user_130953260

When you do bootstrap, you sample your data with replacement, and re-estimate the parameters. In this case you fit the distribution to a dataset that is sampled from your original dataset.

With these N estimates, you get the median, and the confidence interval.

If you want the parameter that best fits your data, you need to get it from fitdist(serving, "gamma")

user_122584590
Related