ivreghdfe: estimates post: matrix has missing values r(504)

[Steffen Mauch]

Hello everyone,

I trying to implement the ivreghdfe command and I'm having trouble to understand the output/the error messages I get.

Here is some example data from my first 20 observations:

* Example generated by -dataex-. To install: ssc install dataex
clear
input float id byte wave long lmonth_profits_noout byte(q35_covid treatment_1) float(interaction1 agent_availability agent_proximity interaction2 interaction3)
 1 1 400000  0 0  0 1 3  0  0
 1 2  85000 12 0  0 1 3 12 36
 2 1  25000  0 1  0 1 3  0  0
 2 2  50000 20 1 20 1 3 20 60
 3 1  50000  0 0  0 1 0  0  0
 3 2      0  1 0  0 1 0  1  0
 4 1 500000  0 0  0 1 0  0  0
 4 2 120000  8 0  0 1 0  8  0
 5 1 400000  0 1  0 1 2  0  0
 5 2 600000  1 1  1 1 2  1  2
 6 1  20000  0 0  0 1 0  0  0
 6 2      0  0 0  0 1 0  0  0
 7 1 200000  0 0  0 1 2  0  0
 7 2 200000 20 0  0 1 2 20 40
 8 1 240000  0 0  0 1 5  0  0
 8 2 250000 12 0  0 1 5 12 60
 9 1  50000  . 0  . 1 2  .  .
 9 2      .  . 0  . 1 2  .  .
10 1 200000  0 1  0 1 1  0  0
10 2 130000  8 1  8 1 1  8  8
end

Î'm running the following command:

ivreghdfe lmonth_profits_noout q35_covid (treatment_1 interaction1 = agent_availability agent_proximity interaction2 interaction3), first absorb(i.id i.wave) cluster(tc_name)

which gives me the following output

(dropped 83 singleton observations)
note: variable #2 is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09)
note: variable #5 is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09)
note: variable #6 is probably collinear with the fixed effects (all partialled-out values are close to zero; tol = 1.0e-09)
(MWFE estimator converged in 2 iterations)
estimates post: matrix has missing values
r(504);

Can somebody explain me what the problem is here? What are "variable #2", "variable #5" etc. refering to?

The problem appears to be the individual-level fixed effects (i.id). If I omit them the regression runs as expected. Is it even possible to include individual-level fixed effects in my case?

[iris lee]

Hi, I met the same problem. Did you solve it? Could you please give me any advice? Thanks a lot!

[Clara Heidelberg]

I would be alos thankfull if anyone has an answer on this.

[Daisuke Adachi]

I also have the same problem. If anybody knows the issue, that would be great.

The only clue I have is that when I removed the `first` option, the command went through. Hence, I suspect it has something to do with computing the test statistics for the first-stage tests.
This is true both for my own data and OP's MRE. I am using Stata 18.0.