Calculating top and bottom 2.5%

[Lindsey Buck]

Hi everyone,

I am trying to add 95% confidence intervals to a kdenstiy plot. I think I need to caculate the top 2.5% and bottom 2.5% of my variable b, but I am not sure how to do this and to add it to the plot as lines (line at lower and upper). Any ideas? Thanks!

preserve
set seed 692525
tempfile temp  // to hold the results from each repetition
permute beta1 b = _b[beta1], saving(`temp') reps(10):reghdfe hous_all_act_ind beta1 gender2 race2 age_at_refer if ym>=`=tm(2018m12)', absorb(i.hh_zip2 i.yearmonth)
su b
use `temp', clear
kdensity b, lwidth(0.3) lcolor(cranberry)  xline(-0.027, lwidth(0.3) lcolor(ebblue) lpattern(longdash)) xlabel (-0.03 (.01) 0.005) xtitle(Estimated Placebo Betas) ytitle(Density) title(Housing Risk Services) graphregion(color(white)) bgcolor(white) text(150 -0.015 "Actual Beta = -0.027", size(small)) 
graph export "placebo1.png", as (png) replace
restore

[Nick Cox]

You've asked this before. This is clearer to me than the earlier version I read, but please go back and mark the earlier threads as solved with a cross-reference to here-- if this one reaches a solution.

If I understand this correctly, your variable b contains a sampling distribution of estimates, so you want its 2.5% and 97.5% percentiles.

Here is an analogue.

. clear

. set seed 2803

. set obs 1000
Number of observations (_N) was 0, now 1,000.

. gen y = rnormal()

. centile y, centile(2.5 97.5)

                                                          Binom. interp.  
    Variable |       Obs  Percentile    Centile        [95% conf. interval]
-------------+-------------------------------------------------------------
           y |     1,000        2.5    -1.90444       -2.032183    -1.77024
             |                 97.5    2.038576        1.889042    2.317331

. return list

scalars:
             r(n_cent) =  2
                  r(N) =  1000
               r(ub_2) =  2.317331126340183
               r(lb_2) =  1.889041952620063
                r(c_2) =  2.038575804233551
               r(ub_1) =  -1.770239873276673
               r(lb_1) =  -2.032183029418558
                r(c_1) =  -1.90443973839283

macros:
           r(centiles) : "2.5 97.5"

The estimates can be read off as r(c_1) and r(c_2).

[Lindsey Buck]

This fixed the problem! Thank you so much for your help. I thought maybe if I asked it different it would make more sense. Thanks so much!