quaids warning

[Helyley SON]

Hi,

I have a questions about the -quaids- command.
I am new to demand system estimation so any help would be absolutely appreciated.

My version of Stata is Stata SE 13.1 on Windows 10. I am trying to estimate a demand system for 12 broad categories of goods, using 10 years of Household Budget Survey data from 200 to 2015. In total, I have 88,715 observations/households. CPI data is used as prices and all household from a particular year face identical prices.

My question is that every time I run the command, quaids, I get "Warning: variance matrix is nonsymmetric or highly singular" and thus no standard errors for the parameters are estimated.


Please let me know, if any of the above is not clear.

Thank you in advance for your time and help

[David Achiles]

please check your demographics variables, which some of them is giving collinearity,, especially age group,,, how do you use age group variable in estimation, as dummy or continues ?

[William Lisowski]

Welcome to Statalist!

Do you have similar results if you omit the vce option? I'm a bit surprised to see your standard errors clustered by year rather than by household, which leads me to wonder if the collinearity is being introduced by the use of clustered errors. But then, I probably know less than you about demand system estimation. I do know that in general, starting simple and adding complexity can help with debugging. Following the line of thought in post #2, you could eliminate the vce and demographics options, and if that solves the problem, add the demographics variables separately in two runs, if each works then add them both in a third run, then add the vce option in a fourth run.

[Juan Caro]

Based on what you are proposing, your problem comes from the fact that prices are unique for each year, therefore there is not enough variability to estimate.

[Kelly Xia]

Could you send the _quaids_utils.mata file, my e-mail address is [email protected]