LISTSERV at the University of Georgia
Menubar Imagemap
Home Browse Manage Request Manuals Register
Previous messageNext messagePrevious in topicNext in topicPrevious by same authorNext by same authorPrevious page (January 2003)Back to main SPSSX-L pageJoin or leave SPSSX-L (or change settings)ReplyPost a new messageSearchProportional fontNon-proportional font
Date:         Tue, 28 Jan 2003 17:48:04 +0100
Reply-To:     Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From:         Asesoría Bioestadística
              <bioestadistica@eresmas.net>
Subject:      Re: Confidence intervals of various percentiles.
Content-Type: text/plain; charset=us-ascii

Hi Margo:

Just for the median (not for other percentiles, sorry)

* If you are using SPSS 11.x:

COMPUTE one = 1 . RATIO STATISTICS amount1 WITH one /PRINT = CIN(95) MEDIAN . RATIO STATISTICS amount2 WITH one /PRINT = CIN(95) MEDIAN .

* Older versions of SPSS:

The following MACRO will do the same:

define cimed95(!positional !tokens(1)) sort cases by !1 (A). matrix. get medvar /var=!1. compute n=nrow(medvar). do if n le 284. + compute lowval=(n+1)/2-0.9789*sqrt(n). + compute low=trunc(lowval). + do if n=17. + compute low=5. + end if. + do if n=67. + compute low=26. + end if. + do if n le 120. + compute pmiss95={0;0;0;0;0;.0313;.0156;.0078;.0391;.0215;.0117;.0386; .0225;.0129;.0352;.0213;.0490;.0309;.0192;.0414;.0266;.0169;.0347;.0227; .0433;.0290;.0192;.0357;.0241;.0428;.0295;.0201;.0351;.0243;.0410;.0288; .0470;.0336;.0237;.0385;.0275;.0436;.0315;.0488;.0357;.0259;.0400;.0293; .0444;.0328;.0489;.0365;.0270;.0402;.0300;.0441;.0331;.0479;.0363;.0273; .0396;.0300;.0430;.0328;.0464;.0356;.0498;.0385;.0295;.0414;.0319;.0444; .0344;.0474;.0369;.0286;.0395;.0308;.0422;.0330;.0448;.0352;.0475;.0375; .0295;.0398;.0314;.0422;.0334;.0446;.0355;.0470;.0375;.0495;.0396;.0316; .0417;.0334;.0439;.0352;.0460;.0371;.0482;.0390;.0313;.0409;.0330;.0428; .0346;.0448;.0363;.0467;.0380;.0487;.0398;.0323;.0415;.0338;.0433;.0353}. + compute realcov=100*(1-pmiss95(n)). + end if. end if. do if n gt 284. + compute low=trunc((n+1)/2-0.98*sqrt(n)). end if. compute upper=n+1-low. compute lower95=medvar(low). compute upper95=medvar(upper). print {lower95,upper95} /format="f10.2" /clabels="Lower 95","Upper 95" /title="95% CI for median". do if n le 120. print realcov /format="f8.1" /title="Exact coverage (%)". end if. end matrix. !enddefine.

* Macro call: process variables one by one.

cimed95 amount1. cimed95 amount2.

HTH

Marta Garcia-Granero

> Hi Listers -- > > Does anyone know how to obtain confidence intervals for various percentiles. > > E.g., if I do the following: > > EXAMINE > VARIABLES=amount1 amount2 > /PLOT NONE > /PERCENTILES(10,25,50,75,90) HAVERAGE > /STATISTICS DESCRIPTIVES > /CINTERVAL 95 > /NOTOTAL. > > I get the mean with confidence intervals -- but not, in the percentile > chart, the median and other percentiles. > > Any thoughts? > > Margo Schlanger > > ______________________ > Margo Schlanger > Assistant Professor of Law > Harvard Law School > Cambridge, MA 02138 > (617) 495-4626 > (253) 295-6089 (fax)


Back to: Top of message | Previous page | Main SPSSX-L page