Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: getting more precise numbers from -summarize-

From   Maarten Buis <>
Subject   Re: st: getting more precise numbers from -summarize-
Date   Thu, 19 May 2011 09:35:57 +0200

On Wed, May 18, 2011 at 10:18 PM, Hewan Bela wrote:
> This is important to me because I copy and paste the summarised results in excel to undertake computations with them, and if the numbers in the summary column are imprecise (e.g. a mean of 1.2e+08  for var4, where I wanted the mean to appear as a non-scientific number) then my excel calculations will be very imprecise as well.

That is almost always wrong. Anything you can do in Excel you can
also/better do in Stata. The real advantage of programs like Stata is
that you can (and must) document your work by doing it all in one (set
of) do-files. Remember that the argument we are making is: We have a
question --> we get data --> we summarize that data (do statistics)
--> we answer our question. The strength of our argument is based on
the fact that the conclusion comes from stuff we have seen (the data),
the summarizing/statistics part is only there to make the necessary
patterns visible. For your argument to retain its strength there needs
to be an exact record of what happens at the summarizing/statistics
stage(*). One test is: do an entire analysis for one paper, don't
touch that work for a month, after a month try to reproduce those
exact same numbers. If you did not store your analysis in a .do file
you will almost certainly fail that test.(**)

If you still insist on degrading the strength of your own argument,
you can look at -help collapse- in combination with -help outsheet-,
with the former you can create a new dataset containing the summary
statistics and with the latter you can export it to something Excel
can read.

Hope this helps,

(*) This is of course a simplification, most of the real damage often
happens at the data preparation/"cleaning" stage, so you also need an
exact record of that too.

(**) This also has practical relevance, as this test is a pretty
accurate/optimistic representation of what happens when you get  the
reviews back from a paper you submitted to some journal...
Maarten L. Buis
Institut fuer Soziologie
Universitaet Tuebingen
Wilhelmstrasse 36
72074 Tuebingen

*   For searches and help try:

© Copyright 1996–2017 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index