Thanks to Kit Baum, there is now a new version of the -parmest- package
downloadable from SSC. Type -ssc desc parmest- to find out more.
The -parmest- package saves the output of a Stata estimation command as a
data set with one observation per parameter, or one observation per
parameter per by-group, and data on estimates, confidence limits, P-values
and other parameter attributes. The new version has been improved in two ways.
First, -parmest- and -parmby- now require the user to specify at least one
of the three options -saving()- and/or -norestore- and/or -fast-. These
options determine whether the output data set is saved to a disk file,
written to memory (overwriting any pre-existing data), or both. Previously,
if the user specified none of the above options, then -parmest- and
-parmby- wrote the output data set to memory, overwriting any existing
data. When I presented -parmest- at the 2002 UK Stata User Meeting, Bill
Gould felt strongly that this default did not do enough to protect users
from their own stupidity, and I decided that he was probably right. I hope
this does not inconvenience existing -parmest- users too much, and I have
supplied what I hope is a very explicit error message in the case where
none of the above options are specified.
Second, the output data set now contains a new variable -dof-, containing
the degrees of freedom used for calculating confidence limits, if these
confidence limits are calculated using a t-distribution. This variable can
have different values for different parameters of the same model, if vector
degrees of freedom are used. Currently, to the best of my knowledge,
official Stata estimation results only specify vector degrees of freedom
for the commands -svymean-, -svyratio- or -svytotal- when the -available-
option is specified. However, this might possibly change in the future,
because vector degrees of freedom are probably a good idea when Huber
sandwich variances are used. If this happens, then it will be easy to adapt
-parmest- to cater for this.
Apart from Kit Baum, I would also like to thank Bill Gould for his very
helpful constructive advice about protecting users from their own
stupidity, and Nick Winter for informing me that -svymean-, -svyratio- and
-svytotal- use vector degrees of freedom when the -available- option is
specified.
Best wishes
Roger
--
Roger Newson
Lecturer in Medical Statistics
Department of Public Health Sciences
King's College London
5th Floor, Capital House
42 Weston Street
London SE1 3QD
United Kingdom
Tel: 020 7848 6648 International +44 20 7848 6648
Fax: 020 7848 6620 International +44 20 7848 6620
or 020 7848 6605 International +44 20 7848 6605
Email: [email protected]
Opinions expressed are those of the author, not the institution.
