Stata Technical Bulletin
13
Syntax
diest Weeghti∖ [if exp [in range] [, {ci ∣mean∣ sezp} level(#) tdf(#) eform(nαme) Iv first
fb(fmt) fse(fmt) fzt(fmt) fp(fmt) fci(fmt) fm(fmt) fsd(fmt) ]
Options
sezp, ci, mean select the display mode. Only one of these options may be specified. sezp is the default.
sezp displays estimates, standard errors, z or t statistics, and 2-sided p-values.
ci displays estimates and confidence intervals.
mean displays estimates, 2-sided р-values, and the mean and standard deviation of the variables.
level (#) specifies the confidence level, in percent, for confidence intervals of coefficients. The default is level (95) or as set
by set level.
tdf (#) specifies the degrees of freedom of the d distribution used to estimate р-values and confidence intervals. Noninteger
values are permitted. tdf (. ) specifies that the normal rather that the d distribution should be used. The column header
shows whether the normal(z) or d distribution is used.
Most estimation commands define the global macro S_E_tdf as the appropriate degrees of freedom for a d distribution,
or to missing if the normal distribution should be used. If the option tdf has not been specified, diest checks whether
S_E_tdf can be used. If S_E_tdf is not available, the normal distribution is used.
eform (name) specifies that the parameter estimates should be exponentiated. In accordance with Stata’s regular behavior, the
standard errors are transformed accordingly; z and s statistics and р-values values are unchanged; and the confidence
interval is exponentiated. The argument of eform specifies the name to be displayed above the column with transformed
coefficients.
Iv specifies that variable labels are displayed (right-aligned) before, instead of after, the variable names.
first specifies that only estimation results pertaining to the first equation are displayed.
Formats
The format of the columns can be specified via options. A format should be a legal Stata display format, though the leading
percentage sign (7,) may be omitted. Examples of display formats: 7,9.4f, 8.3g, and 7,9.Og.
|
Option |
Description |
Default |
Display mode |
|
fb |
parameter estimates |
7,9.0g |
sezp, ci, mean |
|
f se |
standard errors of estimates |
7,9.0g |
sezp |
|
fzt |
o or z statistics |
7,7.3f |
sezp |
|
fp |
p-values |
7,6.3f |
sezp, mean |
|
f ci |
confidence intervals |
7,9.0g |
ci |
|
fm |
mean of variables |
7,8.0g |
mean |
|
f sd |
standard deviation of variables |
7,8.0g |
mean |
Remarks
diest requires that estimation commands post all relevant information for post-estimation commands. Estimates and
estimated covariance matrices are indeed readily available via matrix get. Variable names associated with parameters can
usually be obtained from the names of columns in the matrix of estimates (exception: mlogit). The dependent variables are
usually available via the global macro S_E_depv (there are exceptions; e.g., mvreg). The degrees of freedom for approximate
d distributions of estimates are often made available via the global macro S_E_tdf (there are exceptions; e.g., regress and
ano va).
diest tries to deal with these inconsistencies as far as I was interested in running the exceptional commands myself.
The display mode mean requires that the estimation sample (if, in) and the weighting are available as to post-estimation
commands in order to compute the summary statistics for the right sample and weight. Unfortunately, only a few estimation