ivprobit test of instrument strength

[punit]

Hi All. I am using Stata 13 to run ivprobit, and my question relates to testing the strength of the instrument used. From what I have researched so far, it appears that unlike linear regression (ivreg, ivreg2), there is no test for checking the strength of instrument in ivprobit. Is that correct? Do I need to use ivreg2 for this purpose?

I am aware of, and use, the weak instrument robustness test developed by Finlay, Magnusson and Schaffer, 2013, but I have been advised to report results for test of instrument's strength. Also, if I use the twostep option with ivprobit, would the F-statistics obtained work for this purpose?

Please advise. Thanks a ton in advance.

[Jeff Wooldridge]

In ivprobit the reduced from for the endogenous explanatory variable is linear, so you can use the same diagnostics as in the linear case. So, yes, you can estimate the linear version of the model by ivreg2. Or, as you suggested, just use the F statistic obtained with the two-step option in ivprobit.

As far as I know, no one has actually studied what appropriate rules-of-thumb are for the F statistic when the structural model is nonlinear, but using the Stock-Yogo recommendations is a start.

[punit]

Thanks for your helpful and quick response, Jeff. It's really appreciated. If you can point me to any literature/citations that I could use, that would be doubly helpful. Regards, P.

[Arthur van Eeden]

Hi, also want to show the relevance of my instrument. However I cant do an seperate OLS regression with the first-stage, and corresponding do an F-test. It's so that my "X" is a dummy as well (similair to my Y and instrument). So far I have the following code:

ivprobit debtequity (proxy1or2yearbefore=R1to2 R2to1) MBratio2 saleln roa financialslack rendtosales assettangibility bookleverageratio if inrange(year, 2003, 2007), first

I hope that somebody can tell me how to elaborate on the relevance of my instruments.

Best,

Arthur van Eeden

[Embarika Farouk]

Hello everyone,

I am doing same analysis, I am using ivprobit command with two-stage analysis.
after I run my regression I used the post-estimation (weakiv) to test "the Weak instrument robust tests for IV probit". However I am still searching how to interpret the results I have got through 'wealiv'.

My question to Dr Jeff, please I can understand that the first stage with ivprobit is linear but I did not get it how can I estimate the linear version of the model by "ivreg2". And
use the Stock-Yogo recommendations.

I would appreciate if you clarify that to me please.

Regards

Embarika

[Embarika Farouk]

Hi All. I am using Stata 13 to run ivprobit, and my question relates to testing the strength of the instrument used. From what I have researched so far, it appears that unlike linear regression (ivreg, ivreg2), there is no test for checking the strength of instrument in ivprobit. Is that correct? Do I need to use ivreg2 for this purpose?

I am aware of, and use, the weak instrument robustness test developed by Finlay, Magnusson and Schaffer, 2013, but I have been advised to report results for test of instrument's strength. Also, if I use the twostep option with ivprobit, would the F-statistics obtained work for this purpose?

Please advise. Thanks a ton in advance.

Please refer to this docu., https://www.stata-journal.com/sjpdf....iclenum=st0171

In this doc., they provide a general robust test for weak IV of different classes. They use rivtest command after ivprobit, ivreg2 ....

Hope that hleps.

[Embarika Farouk]

In ivprobit the reduced from for the endogenous explanatory variable is linear, so you can use the same diagnostics as in the linear case. So, yes, you can estimate the linear version of the model by ivreg2. Or, as you suggested, just use the F statistic obtained with the two-step option in ivprobit.

As far as I know, no one has actually studied what appropriate rules-of-thumb are for the F statistic when the structural model is nonlinear, but using the Stock-Yogo recommendations is a start.

Hello
I ma using weakiv as postestimation after IVprobit, still want to understand and interpret the validity tests.

Also my question,please how I can understand that the first stage with ivprobit is linear but I did not get it how can I estimate the linear version of the model by "ivreg2". And
use the Stock-Yogo recommendations.

I would appreciate if you clarify that to me please.

Regards

Embarika