Discussion: View Thread

  • 1.  Criteria for Relative Standard Error in SAE

    Posted 01-29-2016 12:58

    Hello all,

    I have a client from a public sector that produces official statistics from survey and makes them available to public. Like many federal agencies, they have a guidline based on the magnitude of RSE (Relative Standard Error) to determine whether point estimates should be made public, made public with warnings, or suppressed.

    They are planning to start providing information on a smaller geographical area. And RSE criteria being more of an art than science, they want to know if there are accepted criteria (used by any respectable agencies) for RSE specificially for the estimators based on smaller area. That is, do some agencies allow more lenient criteria for small area estimators than estimators based on larger samples?

    Please note that this question is not about different approaches of small area estimation: I am aware of small area estimaton methods that "borrow information" from neighboring small areas and use auxillary predictors and, in turn, reduce RSE altogether. We are looking into that direction but at the same time, they want to know if some agencies have guidelines for reliability specifically suited for small area estimators.

    Thank you in advance for your advice.

    ------------------------------
    Ryung Kim, PhD
    Missional Consulting Group, LLC
    Also Associate Professor at Albert Einstein College of Medicine
    ------------------------------



  • 2.  RE: Criteria for Relative Standard Error in SAE

    Posted 01-30-2016 13:48

    It's not quite the same thing, but some of the large nationally representative surveys conducted by CDC have a rule that you should not report any estimates based on fewer than 30 records and (more relevant to your inquiry) any estimates with a relative standard error larger than 30%. See, for example, Reliability of Estimates.

    ------------------------------
    Stephen Simon, blog.pmean.com
    Independent Statistical Consultant
    P. Mean Consulting



  • 3.  RE: Criteria for Relative Standard Error in SAE

    Posted 02-01-2016 09:24
      |   view attached

    To use the legal term,  I find the "rules" for data suppression to be "arbitrary and capricious."   Klein etal summarize the variety of rules used by surveys associated with the Health People 2010 goals.  I don't know of anything specific to SAEs.

    http://www.cdc.gov/nchs/data/statnt/statnt24.pdf

    I have collected many other references on this perennial topic, and I would be happy to share them with you or others who might be interested. Feel free to contact me directly off list.

    Bob

    ------------------------------
    Bob Gerzoff, MS, PStat®
    Statistician\Team Lead
    Data Analysis and Survey Support
    Epidemiology Branch
    Office on Smoking and Health
    U.S. Centers for Disease Control and Prevention
    rcg8@cdc.gov

    Attachment(s)