Non-Parametric Bootstrapping

[Melissa Crunkhorn]

Hi,
I am using Stata/IC 16.1 and I would like to calculate the 95%CI for a variable (burden) that is the product of two random variables (using an offset) - count and mean duration with an exposure offset.
This dataset represents a count of injuries, their cumulative duration and the individual's exposure duration within the surveillance period.

To account for the uncertainties, I have been recommended to use a non-parametric bootstrapping approach with 1000 iterations as one approach.

Please see a sample of the dataset below.

* Example generated by -dataex-.
clear
input long id double(injured duration exposure)
18829 6 268 1461
18830 6 673 1461
18832 8 378 1461
20896 1 81 1128
20899 4 94 1461
20902 3 448 761
20906 5 202 1461
20909 6 224 1461
20913 9 556 761
20914 7 372 1461
end

To calculate the burden of the population, the following code has been used previously:

*collapse data to summary level
collapse (sum) injured duration exposure
 
*create mean duration variable
gen duration_mean = duration/injured
 
*create incidence variable
gen incidence = injured/exposure*1000
 
*create burden variable
gen burden = incidence*duration_mean

Is there any recommended approach to 1) streamline the approach above, 2) allow <bootstrap> to be ran prior to the collapse?

Many thanks.

[Carlo Lazzaro]

Melissa:
assuming that your calculation is (more or less) right till -burden-, yoummay want to consider something along the following lines:

. . quietly sum burden

. bootstrap r(mean), reps(1000): quietly sum burden

Bootstrap results                                        Number of obs =    10
                                                         Replications  = 1,000

      Command: summarize burden
        _bs_1: r(mean)

------------------------------------------------------------------------------
             |   Observed   Bootstrap                         Normal-based
             | coefficient  std. err.      z    P>|z|     [95% conf. interval]
-------------+----------------------------------------------------------------
       _bs_1 |   290.4472    68.5281     4.24   0.000     156.1346    424.7598
------------------------------------------------------------------------------

. estat bootstrap, all

Bootstrap results                               Number of obs     =         10
                                                Replications      =       1000

      Command: summarize burden
        _bs_1: r(mean)

------------------------------------------------------------------------------
             |    Observed               Bootstrap
             | coefficient       Bias    std. err.  [95% conf. interval]
-------------+----------------------------------------------------------------
       _bs_1 |   290.44722  -1.439478   68.528096    156.1346   424.7598   (N)
             |                                       168.4967   432.1902   (P)
             |                                       177.2014   460.8813  (BC)
------------------------------------------------------------------------------
Key:  N: Normal
      P: Percentile
     BC: Bias-corrected

. 
.