Link to the last RSS article here:
Null Hypothesis Significance Testing
Delivering Results to the End User: Two Stata 8.0
Examples, Part 1
One area in which there
is a great deal of variance across disciplines in the social sciences
is the presentation of results from statistical analyses. I am
comfortable reporting analysis in my own field (political science),
but in my position in Academic Computing Services I am constantly
amazed at the differences in reporting results from discipline to
discipline. In this article I’ll discuss two Stata commands that can
be used to report a variety of parameters, descriptives, and
calculations for delivery to the end user.
You’ve Run Your Models…Now What?
Running statistical models is less than half the battle.
Most statistical software, including Stata 8.0, is automated
to the point where you can point-and-click your way to
whatever model you are attempting to specify. The real works
comes in presenting your results and interpreting them in a
meaningful manner.
The two commands we will overview today are –estimates
table- and –listtex-. Before progressing it would be prudent
to insure that both packages are installed and working on your
version of Stata. This can be checked by typing –help
estimates table- and –help listtex- respectively. If you see
Stata help documentation for these packages after typing these
commands, then these packages are installed on your system. If
you do not see Stata help documentation, then you can find and
install these packages by typing –findit estimates table- and
–findit listtex- respectively from the Stata command line if
your computer is connected to the internet. If you are
connected via a dial-up connection, then you may experience a
slight wait for the packages to download. If you are connected
by a dedicated connection such as a cable modem, DSL, or here
at UNT, then the wait will be negligible.
In the Table and Dreaming: What to Include and How
Depending on the requirements of your original piece of
research, your replication, your thesis, or your dissertation,
what you will want to include in tables and figures. I am
going to use examples in this article reporting coefficients
and standard errors for estimated parameters. Both –estimates
table- and –listtex-
Using –estimates table-
From the Stata 8 help file for –estimates
table-:
Display
one or more estimation results in a table
Basic syntax
estimates table [namelist] [, stats(scalarlist) equations(1)
star ]
Full syntax
estimates table [namelist] [, stats(scalarlist)
equations(matchlist) keep(keeplist) drop(droplist)
star[(#1 #2 #3)] coded b[(fmt)] se[(fmt)] t[(fmt)] p[(fmt)]
stfmt(fmt) eform
varwidth(#) modelwidth(#) label style(style_spec)
newpanel title(str) ]
where
style_spec = oneline | columns | nolines
Description
estimates table displays coefficients, "significance stars",
standard errors, t/z-statistics, and p-values and scalar
statistics of one or more models previously fitted and stored
by estimates store. The last estimated results may be
referenced as a period (.) even if not yet stored. If no
model is specified, estimates table reports about the last
fitted model.
estimates table produces a table in which the coefficients are
matched by the name of equation and coefficient; seeoption
equations() to match equations by number. Each model in
namelist is presented in a column of the table.
estimates table displays blanks in the table cells for those
coefficients that are not in a particular model.
estimates table obeys the linesize.
For this example we will use the
following:
clear;
#delimit ;
set obs 100;
quietly {;
gen x1=uniform(); gen x2=uniform(); gen x3=uniform();
gen y=1+x1+x2+x3+invnorm(uniform());
regress y x1;
estimates store model1;
regress y x1 x2 ; estimates store model2;
regress y x1 x2 x3 ; estimates store model3;
};
tempfile
model1 model2 model3;
estimates
table model1 model2 model3, stats(N chi2) star se;
Using –listtex-
-listtex- is best utilized with some of
Roger Newson’s other Stata user-written packages, all
available using –findit packagename- where packagename is the
name of the package you are trying to locate. For this
particular example, -dsconcat- (a package for concatenating
results into a table) and –parmby- (a by-able version of
–parmest-, another Newson-written package that creates a data
set of parameter estimates) will both need to be added to your
system before you can run the sample code below:
clear;
#delimit ;
set obs 100;
quietly {;
gen x1=uniform(); gen x2=uniform(); gen x3=uniform();
gen y=1+x1+x2+x3+invnorm(uniform());
regress y x1;
estimates store model1;
regress y x1 x2 ; estimates store model2;
regress y x1 x2 x3 ; estimates store model3;
};
tempfile model1 model2 model3;
parmby "reg y x1", lab saving(model1, replace) idn(1)
ids(unadjusted);
parmby "reg y x1 x2", lab saving(model2, replace) idn(2)
ids(unadjusted);
parmby "reg y x1 x2 x3", lab saving(model3, replace) idn(3)
ids(unadjusted);
dsconcat model1 model2 model3;
list parm estimate stderr t min95 max95;
listtex
idnum parm label estimate stderr t min95 max95 using
example1.tex, rstyle(tabular) replace;
In next month’s Benchmarks Online article we will
modify the code presented here using some of the options
available in these packages and evaluate their output.
For More Information:
Newson, Roger. Creating plots and tables of estimation
results using
parmest and friends. Presented at the 8th UK Stata
User Meeting, 20-21 May, 2002:
http://www.kcl-phs.org.uk/rogernewson/usergp/uk2002.pdf
|
