GMM Estimation

[AbuChowdhury]

Hi Experts,

I was trying to do the GMM estimation in proc panel. But I see that AR test result is blank. It shows the following message in the log file:

WARNING: The ARTEST test cannot be calculated for one or more lags due to missing values in the
response and/or explanatory variable.

It shows the same message even if I removed the missing observations for my selected variables. How can I address this issue? is there other ways except proc panel to estimate GMM?

[bobby_sas]

This seems an odd error message.  It may be the case that the problem does not lie with the model regression itself, but from estimation of some auxiliary model used for either starting values or as the basis for a specification test.  If that's the case then the error message is cryptic and could be improved upon, so that it can provide more specific guidance.

If Abu can email me the data (or some subset) and code that reproduces the problem, I'll be glad to take a closer look and give a more specific diagnosis.

[AbuChowdhury]

Hi,

I will send you the partial dataset to your email. Please provide me your email address. My email address is rasel0608@yahoo.com. I will send you the dataset from this email.

[bobby_sas · Accepted solution]

AbuChowdhury emailed me his data, and we were able to ascertain that his missing AR test was caused by gaps in the time series for one or more of the firms in his data.   

[mspak]

Dear bobby_sas,

I also encountered the same problem and receive the following warnings:

WARNING: The ARTEST test cannot be calculated for one or more lags due to missing values in the
response and/or explanatory variable.

However, I am still unsure how to solve the problem by reading the posts in this forum. Could you please share with me as I need to get the AR1 and AR2 test results.

Thank you.

MSPAK

[bobby_sas]

Dear mspak,

I gave the following explanation in another thread:

If the missing values are due to calculations performed by PROC PANEL when generating lagged variables, then this can be alleviated by using the CLAG statement (instead of the LAG statement).  This will replace missing values with cross-sectional means.  Otherwise, you would have to manually exclude the "missing" observations from the analysis by using the WHERE= option. 

--Bobby