Use dplyr to truncate top and bottom percentiles of a numeric variable

Question

I´ve generated a survey-weight. Because outlier survey-weights can lead to very big variances, i´m following a hint from many statistic books: I want to truncate the top 5% and bottom 5% of the survey weight. I would like to use dplyr for this.

#generate data
data<-as.data.frame(cbind(sequence(2000),rnorm(2000,mean=3.16,sd=1.355686))) 
names(data)<-c("id","weight")

#This is how far i got
data2<-data %>% mutate(perc.weight=percent_rank(weight)) %>%
                mutate(perc.weight>0.95 | perc.weight<0.05)

After this, i´ve got two new variables. The first Variable gives percent-ranks of the weights. The second variable shows, if a value exceeds the aimed range.

Now, i want to replace the weights which are in the 95-100 percentile and the weights within the 0-5 percentile with the weight-values that constitute the border of those percentiles.

I would be thankful for any help!

Answer 1

You can use the quantile function togehter with pmin, pmax:

data %>% mutate(weight_trunc = pmin(pmax(weight, quantile(weight, .05)), 
                                          quantile(weight, .95)))