Stata Technical Bulletin
STB-58
dm65.1 Update to a program for saving a model fit as a dataset
Roger Newson, Guy’s, King’s and St Thomas’ School of Medicine, London, UK, [email protected]
Abstract: The parmest command introduced in Newson (1999) has been updated for Stata 6.0.
Keywords: confidence intervals, estimation results.
Introduced in Newson (1999), parmest saves the most recent estimation results to a dataset with one observation per
parameter and is typically used (together with graph) to produce confidence interval plots. It has been updated for Stata 6.0.
Two bugs have been corrected. First, parmest now works on estimation results from multi-equation models (such as those fitted
by mlogit), which previously caused it to fail. Second, when the results are saved using the saving() option, parmest no
longer saves any temporary variables with confusing names. To rule out these bugs and many others, parmest has been tested
extensively using the Stata 6.0 certification script utility (see the on-line help cscript).
Saved results
parmest now saves in r():
Scalars
r(dof) Degrees of freedom for d distribution used for confidence intervals (0 if normal distribution used)
r(nparm) Number of parameters estimated
r(level) Value of level option (confidence level for CIs)
Macros
V (eform) Value of eform option (eform if set, otherwise empty)
Acknowledgment
I would like to thank Vince Wiggins of Stata Corporation for his helpful advice about handling equation names containing
spaces (which are often produced by mlogit for outcomes with value labels).
References
Newson, R. 1999. dm65: A program for saving a model fit as a dataset. Stata Technical Bulletin 49: 2-5. Reprinted in Stata Technical Bulletin
Reprints, vol. 9, pp. 19-23.
dm82 Simulating two- and three-generation families
Jisheng Cui, University of Melbourne, Australia, [email protected]
Abstract: The commands simuped2 and simuped3 for simulating two- and three-generation families are introduced and
illustrated.
Keywords: family data, generations, simulation.
Introduction
Simulation of family data (pedigree) is sometimes required in genetic epidemiology research. Generation of family data using
Stata has advantages over using other computer languages because of the random number generators of probability distributions
built into Stata. Here we present two programs, simuped2 and simuped3. The former is used to simulate two-generation families,
and the latter is used to simulate three-generation families. Figures 1 and 2 give schematic illustrations of the pedigree structure
of the families generated by these programs. A circle represents a female, while a square represents a male. There is a marriage
in the second generation in Figure 2.
(Continued on next page)