[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
Iain Lang <iain.lang@pms.ac.uk> |

To |
"statalist@hsphsun2.harvard.edu" <statalist@hsphsun2.harvard.edu> |

Subject |
st: ?Misreported n of observations when using pweights in Stata 10 |

Date |
Tue, 11 Aug 2009 16:52:35 +0100 |

Dear List Members I am encountering an odd problem when using pweights in regressions in Stata 10. The problem, which I've now come across when using data from two different epidemiological studies (NHANES and the Health and Retirement Study), is that applying pweights seems to alter the number of observations reported as included in the analysis. An example is pasted below. When I saved this dataset in Stata 9 format and opened it there the problem did not replicate - in fact, all the outputs were the same *except* for the number of observations. Given that this has happened to me in two different datasets and in Stata 10 but not Stata 9 I'm guessing that it's something specific to Stata 10. A colleague has encountered something similar. Has anyone else come across this problem? I'm not too worried about this since the estimates, etc., are the same as in Stata 9 but if anyone else has encountered this I'd be interested to hear about it. With regards Iain Example output: . reg hrs06cog35 age gender Source | SS df MS Number of obs = 10478 -------------+------------------------------ F( 2, 10475) = 206.84 Model | 10680.3419 2 5340.17097 Prob > F = 0.0000 Residual | 270441.234 10475 25.8177789 R-squared = 0.0380 -------------+------------------------------ Adj R-squared = 0.0378 Total | 281121.576 10477 26.8322588 Root MSE = 5.0811 ------------------------------------------------------------------------------ hrs06cog35 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age | -.121624 .0061161 -19.89 0.000 -.1336128 -.1096353 gender | .5013875 .1005095 4.99 0.000 .3043697 .6984053 _cons | 29.77283 .4344999 68.52 0.000 28.92113 30.62454 ------------------------------------------------------------------------------ . svyset [pweight=EWGTR] pweight: EWGTR VCE: linearized Single unit: missing Strata 1: <one> SU 1: <observations> FPC 1: <zero> . svy: reg hrs06cog35 age gender (running regress on estimation sample) Survey: Linear regression Number of strata = 1 Number of obs = 27821 Number of PSUs = 27821 Population size = 15217686 Design df = 27820 F( 2, 27819) = 51.63 Prob > F = 0.0000 R-squared = 0.0203 ------------------------------------------------------------------------------ | Linearized hrs06cog35 | Coef. Std. Err. t P>|t| [95% Conf. Interval] -------------+---------------------------------------------------------------- age | -.1533668 .0207634 -7.39 0.000 -.194064 -.1126695 gender | .8502326 .1253616 6.78 0.000 .6045176 1.095948 _cons | 32.94872 1.376334 23.94 0.000 30.25104 35.64641 ------------------------------------------------------------------------------ * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

**Follow-Ups**:**RE: st: ?Misreported n of observations when using pweights in Stata 10***From:*Cameron McIntosh <cnm100@hotmail.com>

- Prev by Date:
**Re: st: Reproducing xtlogit with xtmelogit** - Next by Date:
**Re: st: sampsi and power** - Previous by thread:
**Re st: Chernozhukov and Hong (2002)** - Next by thread:
**RE: st: ?Misreported n of observations when using pweights in Stata 10** - Index(es):

© Copyright 1996–2015 StataCorp LP | Terms of use | Privacy | Contact us | What's new | Site index |