[Nerea Gomez]

Dear Joao Santos Silva ,

First of all, thank you very much for your quick response.

By "it is still loading" I meant that Stata is loading the result, but hours go by and nothing comes out. However, when I try to remove some variables it does work and I get results. That is to say:

1) This works: ivqreg2 Y x1 x2 x3 x4 x5 x6 x7 , inst(Z x2 x3 x4 x5 x6 x7 ) q(.05 .95)

2) This does not work: ivqreg2 Y x1 x2 x3 x4 x5 x6 x7 x8 x9 x10 , inst(Z x2 x3 x4 x5 x6 x7 x8 x9 x10) q(.05 .95)

I tried the mu option, but it still doesn't work. I don't understand why the problem occurs and that's why I was asking if it's normal or not. Regarding the possibility of changing the scale of the dependent variable, it doesn't work for me either.

Thank you again.

Best wishes,

Nerea.

[Joao Santos Silva]

Dear Nerea Gomez,

If you share the data with me, I can try to see if I can make it work and then post the solution here.

Best wishes,

Joao

[Nerea Gomez]

Dear Joao Santos Silva,

Thank you again for your quick response. I have managed to solve it thanks to your previous comment. It turns out that the problem was derived from the scale of one of the explanatory variables, dividing it by 10 I was able to do so without any problem.

Thanks a lot!

Greetings,

Nerea.

[Joao Santos Silva]

Beautiful; thanks for the update.

Joao

[Zarnigor Akhmadalieva]

Dear Professor Joao Santos Silva,

I am now using ivqreg2 for analyzing heterogeneous returns to education.
My model is as follows:

ivqreg2 lwage education experience expersq IQ female married famsize hisp urban urban_child, q(0.1 0.5 0.9) instruments (education father_educ IQ experience expersq female married famsize hisp urban urban_child)

where education is endogenous variable, father_educ is an instrument, and other variables are covariates.
And I want to compare IVQR results with the IV results on the average:

 ivreg2 lwage (education = father_educ) experience expersq IQ female married famsize hisp urban urban_child, gmm2s

The problem is that regular IV regression results are too different from all quantile IV coefficients. Is it because they work differently or am I doing something wrong?

Thank you in advance.

Best regards,
Zarnigor

[Joao Santos Silva]

Dear Zarnigor Akhmadalieva,

You are including education as an instrument; that is not what you want!

Best wishes,

Joao

[Nana Eghan]

Dear Prof Joao Santos Silva ,

Thank you for your works. Please I am trying to estimate the model

ivqreg2 LnExp Age sex Educ hhsize DepR farmsize location dummy_remittance, instruments ( Savings_Acct ) quantile(0.25)

Stata indicates my model is not identified. Please what is wrong with my model specification?

Can I simple estimate without instruments as follows?

ivqreg2 LnExp Age sex Educ hhsize DepR farmsize location dummy_remittance, quantile(0.25)

[Joao Santos Silva]

Dear Nana Eghan,

If you estimate without instruments, you estimate a conditional quantile, similar to what you would get with qreg or qreg2, but imposing the no-crossing condition.

To estimate with instruments, in instruments() you need to list both the instruments and all the exogenous regressors; please see the help file for examples.

Bet wishes,

Joao

[Nana Eghan]

Dear Prof. Joao Santos Silva,

Thank you for your response. I have an endogenous variable in my model hence I need to specify an instrument. I have included the instruments and all exogenous regressors in inst(.), however it is failing to produce results. The STATA keeps running without end.

Please what could be the issue?

Regards,
Mark

[Joao Santos Silva]

Please try with the option "mu".

Best wishes,

Joao

[Nik Anic]

Dear Prof. Joao Santos Silva

Thank you for administering this package and the comments you already made. I am also hitting on an issue I am currently unable to solve. I am running three IV specifications (the subject is irrelevant):

1) ivqreg2 x y z, instruments(y_inst, z_inst) q(0.15 .25 .5 .75 .85)

2) ivqreg2 x y z cov, instruments(y_inst, z_inst cov) q(0.15 .25 .5 .75 .85)

3) local strata i.stratum_reg

- ivqreg2 x y z cov `strata', instruments(y_inst, z_inst cov `strata') q(0.15 .25 .5 .75 .85)

Here, x = dependent variable, y & z = endogeneous regressors, _inst = instruments, cov = the exogeneous covariate(s) (which need to be specified twice according to your comments) and i.stratum_reg = strata fixed effects I include (Running survey population and we have pre-defined strata)

Now, the first two models run smoothly in nearly no time and I get quite reasonable estimates. However, the third faces the same challenge as was posed earlier: It already took over an hour to run and is still computing the results. This is odd, as I do not have an extremely large amount of individual strata. Do you know what might be the reason for this and if / how I could solve this issue quickly? Could it have something to do with using fixed effects in general within this setting (maybe I missed some theoretical conceptions you mentioned).

Thank you very much for your guidance!

Best
Nikolas

[Joao Santos Silva]

Dear Nik Anic,

Does your 3rd specification use 2 sets of fixed effects? How many cases are there in each of them? In any case, please try with the option mu.

Best wishes,

Joao

[Nik Anic]

Dear Joao Santos Silva

Do you mean if I am using 2 different fixed effects? I am only using one fixed effects specification currently, the one for strata. To your second question: there are approx. 80 different strata within the specification (I hope this addresses your point). And how should I add the option mu? should I just place it behind the command of the third specification?

- e.g.: ivqreg2 x y z cov `strata', instruments(y_inst, z_inst cov `strata') q(0.15 .25 .5 .75 .85) mu

I did this and, unfortunately, it still did not compile and is still running.

Best,
Nikolas

[Joao Santos Silva]

Dear Nik Anic,

Sorry, I misread one of your posts above. The estimator is not really designed to deal with fixed effects. You would need many observations per strata for it to be valid. As for the option mu, you are doing it correctly but may not help in this case.

Best wishes,

Joao

[Chhavi Tiwari]

Dear Joao Santos Silva

I am using ivqreg2 command in STATA 16.

The endogenous variable is nnr_dummy and instruments are service proximity and remoteness. I am running the following command:

ivqreg2 lnINCOMEPC nnr_dummy headage age2 lnland caste_dummy2 caste_dummy3 caste_dummy4 Dependent workers incsource_du2 incsource_du3 incsource_du4 incsource_du5 incsource_du6 edu_du2 edu_du3 edu_du4, inst (remoteness headage age2 lnland caste_dummy2 caste_dummy3 caste_dummy4 Dependent workers incsource_du2 incsource_du3 incsource_du4 incsource_du5 incsource_du6 edu_du2 edu_du3 edu_du4) q(.10 .25 .50 .75 .90)

But STATA is not giving any output. I also tried reducing variables in case any variable causing error due to scaling issue. Please suggest.