Strictly, the range referred to here is a range of probabilities,
not that of the corresponding quantiles (values).
Nick
[email protected]
Austin Nichols replied to Yang Lu
If you read the FAQ at
http://www.stata.com/support/faqs/stat/pcrank.html you would see that
this trick imposes symmetry on the extremes of the data. Alternatives
by mi: g r=(srank-srank[1])/srank[_N] if mi==0
by mi: g r=(srank)/srank[_N] if mi==0
would have quantiles in a range [0,1-a] or [a,1] rather than
[a/2,1-a/2].
On Jan 11, 2008 6:11 PM, Yang Lu <[email protected]> wrote:
> Thank you, Austin, the code is very helpful.
>
> One more question (though trivial):
> When calculating the rank, you wrote
> > by mi: g r=(srank-srank[1]/2)/srank[_N] if mi==0
> I noticed that you divided the first srank value by 2 so that the
> percentile for the first obs (with lowest bmi) won't be zero. I'm
> wondering if this is a standard formula, and if not, why can't we just
> write
> by mi: g r=srank/srank[_N] if mi==0
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
