Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: RE: Finding Stata commands
From
Tim <[email protected]>
To
[email protected]
Subject
Re: st: RE: Finding Stata commands
Date
Tue, 05 Oct 2010 00:38:05 +1100
Yes, it is the same question, but it's more a comment. I am asking in
order to get a response (so I guess I'm a troll).
My original post was in reply to a critique of a book. The book was
criticised for, among other things, referring to old Stata commands.
For a review, that is good. I have the latest version of Stata, and I
want information that relates to the current version.
My problem was with the specific statement "most users these days would
look for -psmatch2-".And I wonder why and how most users would look for
that command.
Personally, I would probably read the book, and if I thought the models
could be applied to my problem I would investigate them further (and
perhaps try fitting them). And I would use the commands mentioned in the
book.
And maybe the book is out of date and most users are using a different
command. As I explore further I would probably find the objections to
one command and the reasons to use another command.
But I doubt I (or "most users") would look for -psmatch2-.
On 4/10/2010 23:36, Nick Cox wrote:
Your message has lost its opening, but to be frank this looks, more or less, the same question as before.
I do think you're right to be puzzled.
Kit's comment on -psmatch2- I think in effect presumes that people have been following discussions on how to do propensity score matching in Stata for a while.
That leaves open your reasonable question of how do learners find this out quickly.
As Neil Shepherd and I have already stated, there isn't really such a method, although nothing stops people posting Statalist questions straight out on "What is the best way to do X?" -- and nothing guarantees that they get good answers (although really bad suggestions tend to attract disagreement).
A long while ago, according to legend, Euclid is said to have replied to King Ptolemy's request for an easier way of learning mathematics that there was no royal road to geometry. For "geometry" read "Stata".
Nick
[email protected]
Tim
out why. So here's a summary.
Monitoring Statalist is great for learning what is available (both in
terms of techniques and in Stata commands). But this is a very active
list and I only read posts where the title relates to something I am
working on. I would search the archives if I were trying a new type of
analysis. But I would not have searched for -psmatch2- and I dispute
that "most users" would have.
I don't expect Stata or Statalist to give me the best analysis for my
problem. But if I decide I want to do (for example) propensity
weighting, I would hope -findit- would give appropriate results. And I
still wonder why I have never heard of -psmatch2- when (according to
Statalist) most Stata users interested in propensity matching would look
for -psmatch2-.
On 4/10/2010 21:57, Nick Cox wrote:
As Tim implies, there is an easy and a difficult side to this.
-findit- remains the best single way of finding Stata commands.
. findit propensity score
does point to -psmatch2-, among other materials.
However, this process is dependent on official Stata commands and user-written packages being documented properly, notably by people having previously thought of the keywords you want to type. Sometimes, as in this example, they are fairly obvious, but not always.
Anyone who wants to know what is "best" can only expect only limited help.
1. StataCorp feels free to say which official Stata commands are best for a given purpose (often by downgrading outdated commands in terms of documentation support). They won't grade or evaluate user-written commands, other than by adopting them (which these days is rare).
2. User-programmers should feel free to say which of their own commands are best for a given purpose.
3. You can otherwise best get a sense of what is "best" by following Statalist and listening to what people say -- or do not say. Look for programs that are well supported, well documented and well maintained. Sometimes, a programmer did a good job some years ago and there is no need to update, but often an old, unsupported program may be suspect.
Nick
[email protected]
Tim
I have never used propensity score matching but I considered it recently
and did some reading in the area, and this post attracted my attention.
I have not been following this discussion, but I happened to notice the
claim that "most users these days would look for -psmatch2-".
So my question: why would most users look for -psmatch2-?
Why would anyone look for that name if they did not already know it?
I typed -findit propensity score- and got a lot of results that look
useful. The first few mention -gpscore- (I have not followed links to
know if this is relevant) and -pscore- is mentioned near the end.
In scanning the search results I saw no mention of -psmatch2-. Yet "most
users...would look for psmatch2".
When I think about it, I realise all the Stata commands I use are those
that were mentioned in my training.
If I seriously wanted to do propensity score matching in Stata, how
would I find the correct command?
(and how do I know that -psmatch2- is the correct command?)
More generally, how do I find the best command for an analysis I am
considering but have never used?
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/