estout
Making Regression Tables in Stata
Documentation for eststo
help eststo also see: esttab, estout, estadd, estpost
http://repec.org/bocode/e/estout
-------------------------------------------------------------------------------
Title
eststo -- Store estimates
Syntax
[_]eststo [name] [, options ] [ : estimation_command ]
[_]eststo dir
[_]eststo drop {#|name} [ {#|name} ... ]
[_]eststo clear
options description
--------------------------------------------------------
[no]esample do not/do store e(sample)
title(string) specify a title for the stored set
prefix(prefix) specify a name prefix; default is est
addscalars(...) add scalar statistics
refresh[(#)] overwrite a previously stored set
nocopy clear e() after storing the set
missing use missing values in the by groups
--------------------------------------------------------
by is allowed with eststo if eststo is used as a prefix command, i.e.
specify
by ... : eststo ... : estimation_command
to apply estimation_command to each by group and store an estimation set
for each group; see help by. Note that the implementation of by with
eststo requires estimation_command to follow standard Stata syntax and
allow the if qualifier. Do not use the by ...: eststo: construct with svy
commands.
Description
eststo stores a copy of the active estimation results for later
tabulation. If name is provided, the estimation set is stored under name.
If name is not provided, the estimation set is stored under est#, where #
is a counter for the number of stored estimation sets.
eststo may be used in two ways: Either after fitting a model as in
. regress y x
. eststo
or as a prefix command (see help prefix):
. eststo: regress y x
_eststo is a variant on eststo that, by default, does not store the
estimation sample information contained in e(sample). Essentially,
_eststo is a shortcut to eststo, noesample.
eststo dir displays a list of the stored estimates.
eststo drop drops estimation sets stored by eststo. If name is provided,
the estimation set stored under name is dropped (if * or ? wildcards are
used name, all matching sets are dropped). Alternatively, if # is
provided, the estimation set stored as est# is dropped.
eststo clear drops all estimation sets stored by eststo (and clears
eststo's global macros).
eststo is an alternative to official Stata's estimates store. The main
differences are:
o eststo does not require the user to specify a name for the stored
estimation set.
o eststo may be used as a prefix command (see help prefix).
o eststo provides the possibility to store estimates without the
e(sample) function (either specify the noesample option or use
the _eststo command). Omitting e(sample) saves memory and also
speeds up tabulation programs such as estimates table, estout or
esttab. Warning: Some post-estimation commands may not work with
estimation sets that do not contain the e(sample).
o eststo can add additional scalar statistics to be stored with the
estimation set.
Options
esample causes the information in e(sample) to be stored with the
estimates. This is the default in eststo. Type noesample or use the
_eststo command to omit the e(sample). Note that some post-estimation
commands may not be working correctly with estimation sets that have
been stored without e(sample).
title(string) specifies a title for the stored estimation set.
addscalars(name exp [...] [, replace]) may be used to add additional
results to the e()-scalars of the estimation set before storing it.
Specify the names and values of the scalars in pairs. For example,
addscalars(one 1 two 2) would add e(one) = 1 and e(two) = 2. See
below for an example. The replace suboption permits overwriting
existing e()-returns. Not allowed as names are "b", "V", or "sample".
See estadd for a more sophisticated tool to add additional results to
e()-returns.
prefix(prefix) specifies a custom prefix for the automatic names of the
stored estimation sets. The default prefix is est.
refresh[(#)] may be used to overwrite a previously stored estimation set
instead of storing the estimates under a new name. refresh, specified
without argument, will overwrite the last saved set. Alternatively,
refresh(#) will overwrite the set named est# if it exists. If name is
provided to eststo, existing sets of the same name will always be
overwritten whether or not refresh is specified. refresh() with
argument is not allowed in this case.
nocopy specifies that after the estimation set has been stored, it no
longer be available as the active estimation set.
missing is for use of eststo with the by prefix command and causes
missing values to be treated like any other values in the by
variables. The default is to discard observations with missing values
in the by variables.
Examples
Applying eststo after fiting a model to store the model's results, as in
the following example:
. sysuse auto
(1978 Automobile Data)
. quietly regress price weight
. eststo model1
. quietly regress turn weight foreign
. eststo model2
. estout
--------------------------------------
model1 model2
b b
--------------------------------------
weight 2.044063 .0042183
foreign -1.809802
_cons -6.707353 27.44963
--------------------------------------
Applying eststo as a prefix commmand to fit and store a model in one
step:
. eststo model1: quietly regress price weight
. eststo model2: quietly regress turn weight foreign
. estout
--------------------------------------
model1 model2
b b
--------------------------------------
weight 2.044063 .0042183
foreign -1.809802
_cons -6.707353 27.44963
--------------------------------------
Using eststo with automatic names:
. eststo clear
. eststo: quietly regress price weight
(est1 stored)
. eststo: quietly regress turn weight foreign
(est2 stored)
. estout
--------------------------------------
est1 est2
b b
--------------------------------------
weight 2.044063 .0042183
foreign -1.809802
_cons -6.707353 27.44963
--------------------------------------
Adding ancillary statistics:
. eststo clear
. quietly regress price weight mpg
. test weight = mpg
( 1) weight - mpg = 0
F( 1, 71) = 0.36
Prob > F = 0.5514
. eststo, add(p_diff r(p))
(e(p_diff) = .55138216 added)
(est1 stored)
. estout, stat(p_diff)
-------------------------
est1
b
-------------------------
weight 1.746559
mpg -49.51222
_cons 1946.069
-------------------------
p_diff .5513822
-------------------------
Using the by prefix to store subbroup models:
. eststo clear
. quietly by foreign: eststo: quietly reg price weight mpg
. esttab, label nodepvar nonumber
----------------------------------------------------
Domestic Foreign
----------------------------------------------------
Weight (lbs.) 4.415*** 5.156***
(4.66) (5.85)
Mileage (mpg) 237.7 -19.78
(1.71) (-0.34)
Constant -13285.4* -5065.8
(-2.32) (-1.58)
----------------------------------------------------
Observations 52 22
----------------------------------------------------
t statistics in parentheses
* p<0.05, ** p<0.01, *** p<0.001
Returned results
The name under which an estimation set is stored, is added to the set in
e(_estimates_name).
In addition, eststo maintains two global macros. $eststo contains a list
of the names of the stored estimation sets. $eststo_counter contains the
count of stored estimation sets.
Acknowledgements
Bill Gould suggested to make eststo "byable".
Author
Ben Jann, ETH Zurich, jannb@ethz.ch
Also see
Manual: [R] estimates
Online: help for estimates, esttab, estout, estadd, estpost