ivreg2h "option fe not allowed"

[Kerstin Schmidt]

Hey All,

Only 10 days ago the following command worked perfectly:

ivreg2h y x1 x2 x3 (I = Z), fe cluster(hhid) first

When running it today, it shows me the following error:

option fe not allowed
r(198);

I already dropped the relevant installation files and installed them manually:

cap ado uninstall ftools
cap ado uninstall reghdfe
cap ado uninstall ivreg2
cap ado uninstall ivreghdfe

net install ftools, from("https://raw.githubusercontent.com/sergiocorreia/ftools/master/src/") replace
net install reghdfe, from("https://raw.githubusercontent.com/sergiocorreia/reghdfe/master/src/") replace
ssc install ivreg2, replace
net install ivreghdfe, from("https://raw.githubusercontent.com/sergiocorreia/ivreghdfe/master/src/") replace

... but without any success!

What happened and how can I get the code running again? I have the same problem in Stata 17.0 and 18.5.

Thanks!!!

[Carlo Lazzaro]

Kerstin:
is your Stata fully updated?

[Kerstin Schmidt]

Carlo Lazzaro: Yes, it is. What can I do?

[Carlo Lazzaro]

Kerstin:
what if you go:

. webuse grunfeld 
. ssc install center  
. ssc install bcuse 
. ivreg2hfe invest L(1/2).kstock (mvalue=)

[Kerstin Schmidt]

Carlo Lazzaro: Then it works. Why? How can I adjust my command to produce results?

[Carlo Lazzaro]

Kerstin:
I would go along the following lines (cluster-robust standard errors is not allowed):

. webuse grunfeld 

. . ivreg2hfe invest kstock (mvalue= time ), first

[Kerstin Schmidt]

Thank you Carlo Lazzaro!

I don't quite understand why invreg2h does not run with fixed effects any more but I need to replicate my former command to check additional tests.

The original command is:

 
 ivreg2h y x1 x2 x3 (I = Z), fe cluster(hhid) first

If using "ivreg2hfe" instead, I cannot include cluster-robust errors (as you say). This would not fulfill my aim.

How can I get Stata to run the original code or find an alternative producing the same results?

[Carlo Lazzaro]

Kerstin:
taking a second look at -ivreg2hfe-helpfile, I've found out that the -cluster()- option actually returns cluster robust standard errors.

. ivreg2hfe invest kstock (mvalue= time ), first cluster( company )