[R] Bootstrapped Regression

Rui Barradas ruipbarradas at sapo.pt
Sun Oct 15 09:25:06 CEST 2017


Hello,

Much clearer now, thanks.
It's a matter of changing the function boot calls to return the 
predicted values at the point of interess, education = 50, income = 75.

I have changed the way the function uses the indices a bit, the result 
is the same, it's just the way I usually do it.

pred.duncan.function <- function(data, indices) {
     mod <- lm(prestige ~ education + income, data = data[indices, ])
     new <- data.frame(education = 50, income = 75)
     predict(mod, newdata = new)
}

set.seed(94)    # make the results reproducible

Predicted <- boot(Duncan, pred.duncan.function , 1000)
head(Predicted)
Predicted$t0
boot.ci(Predicted, index = 1, conf = 0.95, type=c("basic", "norm", 
"perc", "bca"))


Hope this helps,

Rui Barradas

Em 15-10-2017 02:22, Janh Anni escreveu:
> Hello Rui,
>
> Thanks for your helpful suggestions.  Just for illustration, let's use the
> well known Duncan dataset of prestige vs education + income that is
> contained in the "car" package.  Suppose I wish to use boot function to
> bootstrap a linear regression of prestige ~ education + income and use the
> following script:
>
> duncan.function <- function(data, indices) {data = data[indices,]
>
> mod <- lm(prestige ~ education + income, data=data,)
>
> coefficients(mod)}
>
> Results <- boot(Duncan, duncan.function , 1000)
> Results
>
> So the 1000 bootstrapped coefficients are contained in Results and I can
> use the boot.ci function in the same boot package to obtain the confidence
> intervals for the, say, education coefficient with something like:
>
> boot.ci(Results, index=2, conf = 0.95, type=c("basic", "norm", "perc",
> "bca"))
>
> Then, suppose I am interested in getting a confidence interval for the
> predicted  prestige at, say, education = 50 and income = 75.  The question
> is how do I get boot to compute 1000 values of the predicted prestige at
> education = 50 and income = 75, so that I can subsequently (hopefully) have
> boot.ci compute the confidence intervals as it did for the bootstrapped
> coefficients? As for prediction intervals, it wouldn't seem conceptually
> feasible in this context?  Thanks again for all your help.
>
> Janh
>
> On Sat, Oct 14, 2017 at 11:12 AM, Bert Gunter <bgunter.4567 at gmail.com>
> wrote:
>
>> R-help is not a free coding service. We expect users to make the effort to
>> learn R and *may* provide help when they get stuck. Pay a local R
>> programmer if you do not wish to make such an effort.
>>
>> Cheers,
>> Bert
>>
>>
>> On Oct 14, 2017 7:58 AM, "Janh Anni" <annijanh at gmail.com> wrote:
>>
>> Greetings!
>>
>> We are trying to obtain confidence and prediction intervals for a predicted
>> Y value from bootstrapped linear regression using the boot function. Does
>> anyone know how to code it?  Greatly appreciated.
>>
>> Janh
>>
>>          [[alternative HTML version deleted]]
>>
>> ______________________________________________
>> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
>> https://stat.ethz.ch/mailman/listinfo/r-help
>> PLEASE do read the posting guide http://www.R-project.org/posti
>> ng-guide.html
>> and provide commented, minimal, self-contained, reproducible code.
>>
>>
>>
>
> 	[[alternative HTML version deleted]]
>
> ______________________________________________
> R-help at r-project.org mailing list -- To UNSUBSCRIBE and more, see
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code.
>



More information about the R-help mailing list