Title
ci_marg_mu -- Simulation-based confidence intervals for predicted marginal probabilities, etc., using gllapred
Syntax
ci_marg_mu lower upper [if] [in] [, options]
options Description ------------------------------------------------------------------------- level(#) set confidence level; default is level(95) reps(#) set number of simulations (must be multiple of 200); default is 1000 dots display a dot for each simulation ------------------------------------------------------------------------- where lower and upper are the names of new variables in which th lower and upper confidence limits will be stored.
Description
ci_marg_mu produces simulation-based confidence intervals for predictions using gllapred varname, mu marg after estimation using gllamm. It repeatedly draws a sample of model parameter values from the estimated asymptotic sampling distribution (i.e., a multivariate normal distribution with mean given by the etimates in e(b) and covariance matrix in e(V)) and obtains predictions using these simulated parameters. It returns the appropriate percentiles in lower and upper. For example, with the level(95) and reps(1000) options, the 25th largest prediction is returned in lower and the 976th largest prediction is returned in upper.
Options
level(#) specifies the confidence level, as a percentage, for confidence intervals. The default is level(95) or as set by set level.
reps(#) specifies the number of simulations to be used. This must be a multiple of 200. The default is 1000.
dots specifies that a dot should be displayed after each simulation to help guage how long the program will run.
Examples
--------------------------------------------------------------------------- Setup . webuse bangladesh
Random-intercept model, analogous to xtlogit . gllamm c_use urban age child*, i(district)
Predict marginal probability for observations where urban = 1 . gllapred prob if urban==1, marg mu
Obtain 95% confidence limits for probability . ci_marg_mu lower95 upper95 if urban==1, level(95) reps(1000) dots
Random-intercept and random coefficient model, correlated random effects, analogous to xtmelogit . generate cons=1 . eq inter: cons . eq slope: urban . gllamm c_use urban age child*, i(district) nrf(2) eqs(inter slope) link(logit) family(binom) adapt ip(m) nip(11) Predict marginal probability for observations where urban = 1 . gllapred prob_rc if urban==1, marg mu
Obtain 95% approximate confidence limits for prediction . ci_marg_mu lower_rc upper_rc if urban==1, level(95) reps(1000) dots --------------------------------------------------------------------------- Setup . webuse lowbirth . generate id = _n
Ordinary logistic regression, analogous to logit . gllamm low age lwt race2 race3 smoke ptd ht ui, i(id) link(logit) family(binom) init
Predicted probabilities, analogous to predict, pr after logit . gllapred prob, marg mu
Obtain approximate 95% confidence limits for probability . ci_marg_mu l u ---------------------------------------------------------------------------
Webpage http://www.gllamm.org
Autor Sophia Rabe-Hesketh
References
Rabe-Hesketh, S., Skrondal, A. and Pickles, A. (2002). Reliable estimation of generalized linear mixed models using adaptive quadrature. The Stata Journal 2 (1), 1-21.
Rabe-Hesketh, S., Skrondal, A. and Pickles, A. (2004). GLLAMM Manual. U.C. Berkeley Division of Biostatistics Working Paper Series. Working Paper 160.
Rabe-Hesketh, S., Skrondal, A. and Pickles, A. (2005). Maximum likelihood estimation of limited and discrete dependent variable models with nested random effects. Journal of Econometrics 128 (2), 301-323.
Rabe-Hesketh, S., Skrondal, A. (2008). Multilevel and Longidutinal Modeling Using Stata (Second Edition). College Station, TX: Stata Press.
Saved results
There are no saved results
Also see
Online: gllamm, gllapred