* quietly {
  /*
     The hat value h{i} is a common measure of leverage in regression.
       The value of h{i} indicates the weight of the contribution of
       Y{i} to the fitted value Y-hat{i}.  This value summarizes the
       potential influence (the leverage) of all Y{i}'s.

                 h{i} is bounded by: 1/n <= h{i} <= 1
                 h-bar = (j+1)/n

     Generally you can use 2(h-bar) as a cutoff for extreme values.
       in small samples this may nominate too many values and you
       can use 2(h-bar).
  */

  regress  y x 
  predict hat, leverage

  sum hat
  local hatcrit =2*r(mean)
  local hatcrit1=round(`hatcrit',.001)

  #delimit ;
  graph twoway (scatter hat  obs),
     ylabel(0(.1).4) ytick(0(.02).4)
     xlabel(1(1)10)
     yline(`hatcrit', lp(dash) lw(thick))
     note("2*hat-bar = `hatcrit1'")
     name(hat, replace)
  ;
  #delimit cr


  /* List Extreme Values of h{i} */
  noisily list obs y x if hat > `hatcrit'

  drop hat
* }