[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

From |
Susan Donath <donaths@cryptic.rch.unimelb.edu.au> |

To |
statalist@hsphsun2.harvard.edu |

Subject |
Re: st: replicate weights using svr commands in Stata8 |

Date |
Mon, 18 Aug 2003 13:47:34 +1000 |

Thanks Nick. This seems to give the correct answer. SD At 06:26 PM 15/08/2003 -0400, you wrote:

Susan DonathAt 12:49 PM 8/15/2003 +1000, Susan Donath wrote:I have a query about the svr programs.As the author of the -svr- routines, I'll take a crack at this one. It sounds from the formula given that these are JK1 (ie, delete-one jackknife) weights, and you should, therefore, use that method. Generally the degrees of freedom for the JK1 method is one less than the number of replicates, or 29 in your case.

I am using a dataset containing unit record data from the Australian Bureau of Statistics (ABS). The ABS provides a set of 30 replicate weights for each case in the dataset. According to their documentation, the estimated standard error of an estimate can be found using the following formula :

SE(y) = sqrt( 29/30 S ( y(g) - y )*( y(g) - y ) )

where y(g) is the gth estimate of the parameter y using the gth set of replicate weights

and y is the estimate of the parameter using the full person weight

The supplied dataset includes a person weight for each case, but no information on strata or psu.

I assume that I can use the svr programs to calculate standard errors but I can't quite figure out the exact syntax of the command. I assume that I need to use the brr version of svrset, but this requires me to enter a number of degrees of freedom, and I don't understand what this refers to.

So, you should use -svrset- to specify the method (jk1), the variable names for the full-sample and replicate weights, and the degrees of freedom:

. svrset set meth jk1

. svrset set pw [sample-weight-spec]

. svrset set rw [replicate-weight-spec]

. svrset set dof 29

Then you can be off and running with your analyses, using the suite of -svr- commands. (By the way, -help survwgt- gives the methods and formulae for variance estimation for each of the various replication methods.)

-Nick Winter

(note the new email address; different from that in the help file for -svr-)

-------------------------------------------------------

Nicholas Winter 308 White Hall

Assistant Professor t:607.255.8819

Department of Government f:607.255.4530

Cornell University nw53@cornell.edu

Ithaca, NY 14853-7901 falcon.arts.cornell.edu/nw53

*

* 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/

Senior Research Fellow

Clinical Epidemiology and Biostatistics Unit

Murdoch Children's Research Institute & University of Melbourne Department of Paediatrics

Royal Children's Hospital

Parkville, Vic. 3052. AUSTRALIA

E-mail: donaths@cryptic.rch.unimelb.edu.au

Tel. 9345 4742 Fax 9345 6000

All numbers area code 03 or international prefix +6

*

* 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/

**References**:**st: replicate weights using svr commands in Stata8***From:*Susan Donath <donaths@cryptic.rch.unimelb.edu.au>

**Re: st: replicate weights using svr commands in Stata8***From:*Nicholas Winter <nw53@cornell.edu>

- Prev by Date:
**Re: st: LR test and robust SE under logit** - Next by Date:
**st: Predicted probabilities for different x values after xtlogit, fe** - Previous by thread:
**Re: st: replicate weights using svr commands in Stata8** - Next by thread:
**st: heteroskedasticity and ordered probit/logit** - Index(es):

© Copyright 1996–2014 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |