Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Gitit Kadar Satat <G.Kadar-Satat@sms.ed.ac.uk> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: gllamm for binary outcome - interpretation? |
Date | Tue, 08 May 2012 19:02:33 +0100 |
Hello Statalist helpers ! (apologies for re-posting, not sure the first time has been received)I ran a gllamm multilevel model to explore a binary outcome. the data I'm using are individuals clustered in schools. The outcome of interest is whether or not they have passed an exam.
I started by estimating an "unconditional model" with no explanatory variable and then ran additional model with the explanatory variables included.
Based on the output (below), how can I estimate the percentage of variance explained at the "grouping" level (that is - schools)? Or is my output missing important information? In other words - I would like to be able to report the % variance in the outcome (whether or not individuals have passed the exam) that is explained by differences across schools. Is this possible?
Here is the main output: *Model 1 (unconditional):. gllamm pass_exam, i(sptn00) pweight(pwt) link(logit) family(binomial) nip(30) adapt
number of level 1 units = 12552 number of level 2 units = 335 Condition Number = 1.1434506 gllamm model log likelihood = -8160.5935 Robust standard errors --------------------------------------------------------------------------------------pass_exam | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------------+----------------------------------------------------------------_cons | -1.31337 .0397836 -33.01 0.000 -1.391345 -1.235396
-------------------------------------------------------------------------------------- Variances and covariances of random effects ------------------------------------------------------------------------------ ***level 2 (sptn00) var(1): .2054349 (.03404487) ------------------------------------------------------------------------------ *Model 2 (explanatories included). xi: gllamm pass_exam i.S4_Parental_nssec0 i.S4_Parental_nvq i.S4_Income_quartiles ,
i(sptn00) pweight(pwt) link(logit) family(binomial) nip(30) adapt number of level 1 units = 12379 number of level 2 units = 335 Condition Number = 281.7895 gllamm model log likelihood = -7870.4423 Robust standard errors --------------------------------------------------------------------------------------pass_exam | Coef. Std. Err. z P>|z| [95% Conf. Interval]
---------------------+----------------------------------------------------------------_IS4_Parent_2 | .1579462 .1034318 1.53 0.127 -.0447763 .3606688 _IS4_Parent_3 | .2136482 .0943537 2.26 0.024 .0287183 .3985781 _IS4_Parent_4 | .205371 .1163226 1.77 0.077 -.022617 .433359 _IS4_Parenta2 | .2970159 .2201864 1.35 0.177 -.1345416 .7285733 _IS4_Parenta3 | .329872 .191429 1.72 0.085 -.0453219 .7050659 _IS4_Parenta4 | .3068724 .2005867 1.53 0.126 -.0862702 .700015 _IS4_Income_2 | .090987 .1359474 0.67 0.503 -.1754649 .3574389 _IS4_Income_3 | .127511 .1269022 1.00 0.315 -.1212128 .3762348 _IS4_Income_4 | .3252327 .1420127 2.29 0.022 .046893 .6035724 _cons | -2.080896 .1809789 -11.50 0.000 -2.435609 -1.726184
-------------------------------------------------------------------------------------- Variances and covariances of random effects ------------------------------------------------------------------------------ ***level 2 (sptn00) var(1): .20224494 (.03582371) ------------------------------------------------------------------------------ Thanks in advance, Gitit -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336. * * 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/