Comments on: About http://magazine.amstat.org Thu, 16 May 2013 21:20:41 +0000 hourly 1 http://wordpress.org/?v=3.5.1 By: Ron Wasserstein http://magazine.amstat.org/about/comment-page-1/#comment-9568 Ron Wasserstein Wed, 08 Feb 2012 19:32:29 +0000 http://magazine.amstat.org/?page_id=81#comment-9568 Lutz,
Your point is well taken, and if you can point me to advice about how to do this better, I’d appreciate it. I’m told we’ve tried to have twice the number of data points needed to estimate the top quantile, but I’m not up on the literature on this subject. We’d welcome your advice for future surveys.
Ron Wasserstein
Executive Director, ASA

]]> By: Lutz Edler http://magazine.amstat.org/about/comment-page-1/#comment-9549 Lutz Edler Tue, 07 Feb 2012 12:32:27 +0000 http://magazine.amstat.org/?page_id=81#comment-9549 The Amer Stat Assoc salary survey calculated 90% percentiles in January 2012 amstat explaining:

“The 90th percentile is provided for any category with 19 or more respondents. That is the minimum sample size for the 90th is n=19.”

Could you provide a justification for this cut off of 19 and what is the general rule and its assumptions used therefore.
There are some papers on such minimal sample sizes for quantiles around. But I think there is no general agreement among statisticians from which (small)sample size on qunatiles should not reported to avoid misuse.

LUTZ EDLER

]]>