An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20110204/30001fc5/attachment.pl>
Extracting specific samples from MCMCglmm
6 messages · Robin Jeffries, Ben Bolker, Jarrod Hadfield
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20110204/820a02b0/attachment.pl>
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I believe that if m is a an MCMCglmm fit, m$Liab will get you the "Posterior distribution of latent variables" (see ?MCMCglmm); I'm not sure whether there is useful information about the structure encoded in the object, but see if colnames(m$Liab) gets you something helpful. Ben Bolker
On 02/04/2011 02:50 PM, Robin Jeffries wrote:
Or alternatively, can I get the samples of the individual random effects out of MCMC glmm?
On Fri, Feb 4, 2011 at 11:01 AM, Robin Jeffries <rjeffries at ucla.edu> wrote:
Hello, I'm very new at using MCMCglmm, so my apologies if my question is rudimentary. I can't even create a sufficient toy example to assist my explanation of what I'm trying to achieve. I've been working in lmer on this for a little while, but since my end result is prediction for specific levels of my random effects, L-serv consensus seems to be that I switch to MCMCglmm. I'm running prediction models on some gaussian (and separately binomial) outcome where I have random intercepts and slopes - within factors. random=~ us(1+slope1 + slope2):Factor1 + us(1+slope1):Factor 2 + Factor3 The short question is how do I create prediction intervals for specific levels of Factors 1-3? Robin Jeffries MS, DrPH Candidate Department of Biostatistics UCLA 530-624-0428
[[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAk1MWmMACgkQc5UpGjwzenMwQQCeKlhelTFBDkeilyB+hb7aJ+HV FTUAoI5BeDVfdcMW+UpkJhnGlC3hDc68 =y+W5 -----END PGP SIGNATURE-----
Hi, If you specify pr=TRUE in the call to MCMCglmm the posterior distribution of random effects will be stored. They are in m$Sol in columns after the fixed effects. You can obtain your predictions "by hand" by extracting the relevant random effects. Alternatively you can use the predict function, but currently this will only predict the data points used in model fitting. In your case you want to use the random effects in the prediction rather than marginalising them so specify marginal=NULL in the call to predict. To obtain prediction intervals instead of confidence intervals specify interval="prediction" in the call to predict. This obtains prediction intervals using posterior predictie simulation so can be slow. Cheers, Jarrod Quoting Ben Bolker <bbolker at gmail.com>:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 I believe that if m is a an MCMCglmm fit, m$Liab will get you the "Posterior distribution of latent variables" (see ?MCMCglmm); I'm not sure whether there is useful information about the structure encoded in the object, but see if colnames(m$Liab) gets you something helpful. Ben Bolker On 02/04/2011 02:50 PM, Robin Jeffries wrote:
Or alternatively, can I get the samples of the individual random effects out of MCMC glmm?
On Fri, Feb 4, 2011 at 11:01 AM, Robin Jeffries <rjeffries at ucla.edu> wrote:
Hello, I'm very new at using MCMCglmm, so my apologies if my question is rudimentary. I can't even create a sufficient toy example to assist my explanation of what I'm trying to achieve. I've been working in lmer on this for a little while, but since my end result is prediction for specific levels of my random effects, L-serv consensus seems to be that I switch to MCMCglmm. I'm running prediction models on some gaussian (and separately binomial) outcome where I have random intercepts and slopes - within factors. random=~ us(1+slope1 + slope2):Factor1 + us(1+slope1):Factor 2 + Factor3 The short question is how do I create prediction intervals for specific levels of Factors 1-3? Robin Jeffries MS, DrPH Candidate Department of Biostatistics UCLA 530-624-0428
[[alternative HTML version deleted]]
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
-----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.10 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAk1MWmMACgkQc5UpGjwzenMwQQCeKlhelTFBDkeilyB+hb7aJ+HV FTUAoI5BeDVfdcMW+UpkJhnGlC3hDc68 =y+W5 -----END PGP SIGNATURE-----
_______________________________________________ R-sig-mixed-models at r-project.org mailing list https://stat.ethz.ch/mailman/listinfo/r-sig-mixed-models
The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20110204/778d75e4/attachment.pl>
An embedded and charset-unspecified text was scrubbed... Name: not available URL: <https://stat.ethz.ch/pipermail/r-sig-mixed-models/attachments/20110204/d1e27493/attachment.pl>