st: RE: RE: RE: RE: IV Command and Reported Results Problem

[Nick Cox <n.j.cox@durham.ac.uk>]

Fine, and many thanks for the clarification. 

The StataCorp position is that the official command -ivregress- supersedes the official command -ivreg-, so there is nothing controversial there in any sense. The authors of both -a- and -b- are naturally always totally entitled to say which they consider better. 

My concern was purely about -ivreg2-, which is the only user-written program mentioned here. 

Others more knowledgeable may be able to say more about your underlying questions, but "your model may be too complicated" is already twice on offer. 

Nick 
n.j.cox@durham.ac.uk 

Sinan Hastorun

OK Nick, in the light of what you have just said, let me say this once and for all: Perhaps I expressed myself poorly or miscommunicated the problem I was having, but I did not intend in any way to disparage anybody's work/ program or assert anything to that effect. I am sorry if anything I've written led to that impression or result.  

(0) I said I am not getting any results reported for those 3 measures.

(1) The STATA press book I used as a guide Microeconometrics Using Stata by Cameroni & Trivedi recommends using ivregress as opposed to ivreg. That is why I was using that. I asked: is it true that it performs better in all cases?

(2) What I meant by "command as a possible source of the problem" is: the command line as I've written it may be the source of the problem I'm encountering. As in ivregress or any other command not being compatible with the model as I've specified it.

I hope this will help alleviate any concerns you may have regarding what I am trying to accomplish here. I will be clearer and more detailed in the future. Of course, you're entitled to your own opinion. Yes the quotation is there and everyone can judge for themselves.

Nick Cox

I am very glad to hear that, but the record shows that 

(0) you reported that you were getting poor results 

(1) you asserted that -ivreg2- is inferior to -ivregress- 

(2) you asked whether "the command may be part of the problem"

-- and I found the conjunction of these statements strange, as I said. 

And despite your explanation I am not impressed by your willingness to make public statements with disparaging implications about users' programs, giving no argument or evidence whatsoever. The quotation is there, below, and the full context makes no difference to what you said. As anyone can find out, I am not an author and am not defending my own work. 

OK, let's move on: You want a comment on, among other things, what may be wrong with the command line as you've written it. Note that as you've only just given it no one was able to comment before now. 

Reminder please: The Statalist FAQ asks explicitly that  

"Say exactly what you typed and exactly what Stata typed (or did) in response. N.B. exactly!"

Nick 
n.j.cox@durham.ac.uk 

Sinan Hastorun

Nick, I am only trying to figure out why I am unable to get these 3 values reported once I run ivregress. I did not imply in any way that theirs or anybody else's programs are to blame for the omission. I am seeking to find out what may be wrong with my model specification or the command line as I've written it. Why would this be a strange question? Stata users post questions about their problems all the time when they get what seem to be unexpected results or in my case, no values reported for certain measures. 

My model contains 51 dummy variables, one for each region represented, that is why the number of explanatory variables appear as 53. So my command is as follows for ivregress:

ivregress 2sls y i1 i2 i3 i4 i5 i6 i7 i8 i9 i10 i11 i12 i13 i14 i15 i16 i17 i18 i19 i20 i21 i22 i23 i24 i25 i26 i27 i28 i29 i30 i31 i32 i33 i34 i35 i36 i37 i38 i39 i40 i41 i42 i43 i44 i45 i46 i47 i48 i49 i50 i51 (x1 x2 = v1 v2 v3 v4 v5 v6), first noconstant

Finally, ivreg2 gives me the R-squared when I use a similar command structure.
 
Nick Cox

It's a strange question to send to Statalist. Two out of three of the very distinguished authors of -ivreg2- are active on this list, and you seem to be asking: I get puzzling and disappointing results. Is your program to blame? 

What do you expect them to say? 

My only guess is that you are trying to fit a model that is too complicated or otherwise unsuitable for your data. As Maarten Buis often advises, retreat to a much simpler model, complicate slowly, and see when and where problems start biting.

But I am not an econometrician. The IV people on this list are better placed to advise, if only on what you should be telling them that you are not. 

Sinan Hastorun

Thank you Nick. I just included that note in my question in order to inquire about whether the command may be part of the problem here. Do you have a take on why I am not getting any actual numerical results reported for Wald chi2(53), Prob > chi2, R-squared?

Nick Cox

Short answer is No; no such simplistic statement deserves assent. 

Enormously longer answers from the experts....

Sinan Hastorun

I am running IV in STATA by using the command ivregress because it is said to perform better than ivreg or ivreg2. Is that correct?


*
*   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/