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

From |
"Kim, Jinseok" <jinskim@mail.utexas.edu> |

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

Subject |
st: test random variance in gllamm |

Date |
Mon, 8 Mar 2004 22:20:58 -0600 |

Hello statalisters, I am using gllamm to conduct a multilevel logistic regression and have a few questions regarding interpretations of test of variance of randome coefficients. I use "allc" option as recommended by gllamm manual (p33-34) and it gave me the standard deviation estimates and their standard errors. Part of stata output is attached here: ***level 2 (nhood) var(1): .81944998 (.44280036) ------------------------------------------------------------------------------ ------------------------------------------------------------------------------ d_phsabs | Coef. Std. Err. z P>|z| [95% Conf. Interval] -------------+---------------------------------------------------------------- d_phsabs | _cons | -2.838067 .202323 -14.03 0.000 -3.234613 -2.441521 -------------+---------------------------------------------------------------- nhoo1 | _cons | .9052348 .2445776 3.70 0.000 .4258714 1.384598 ------------------------------------------------------------------------------ Here's my first question. I understand var(1)=.819=.905^2 from the table but why we need the standard deviation estimate and its standard error to test the significance of random coefficient (random intercept in this case) while we have the variance estimate and its standard error (i.e., var(1): .81944998 (.44280036)), which we could have used for the same test. Is there any reason we have to do the test using standard deviation estimate instead of variance estimate of the same random coefficient? My second question along this line is why the test results are different from each other, which was the case in the gllamm manual example too? In my study, the test result of standard deviation estimate was significant at the 5% alpha level while the same of variance estimate was not. I also understand the multilevel literature recommend the use of likelihood ratio test for the test of random effects but I am just curious. Please help. Thanks much. Jinseok +++++++++++++++++++++++++++++++++++++++++ Kim, Jinseok MS, MSSW Assistant Instructor & Doctoral Candidate at the School of Social Work, University of Texas at Austin E-mail: jinskim@mail.utexas.edu, praxis6887@yahoo.com +++++++++++++++++++++++++++++++++++++++++ * * 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/

- Prev by Date:
**Re: st: p weights in other commnads** - Next by Date:
**st: About sureg** - Previous by thread:
**RE: st: Problems with downloading the collinearity program from ADS** - Next by thread:
**st: About sureg** - Index(es):

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