Estimates post: matrix has missing values! for ivreg2

[Leo wong]

Hi, guys!

I received an error message "estimates post: matrix has missing values" when i'm using ivreg2. I've checked that there's no missing values in the y and controls. I really don't know how to handle this issue, any ideas are appreciated!!

Here's my stata codes:

local x1 male age agesq tmperiod
local x2 s_age
local x3 h_dfct

xi: ivreg2 y (x = iv) `x1' i.educ i.employer2 i.industry i.marea `x2' i.s_educ `x3' i.p_prefect [pw = w_l], cluster(i.home_pref i.p_prefect) partial(i.p_prefect) first

Best!
Leo

[Leo wong]

One more thing, ivreg2 runs successfully if i remove the option part "cluster(i.home_pref i.p_prefect)". No error shows, but it's not the result i want. I need the clsuster for the final result.

[Phil Bromiley]

You don't want the i. in the cluster statement. While xi is not needed for recent Stata versions, I don't know about user-written ivreg2.

[Phil Bromiley]

You'll increase your chances of a useful answer by following the FAQ on asking questions - provide Stata code in code delimiters, readable Stata output, and sample data using dataex. And simplify what you post - you just make it harder for the reader by defining three locals that are just variable names.

[Leo wong]

You don't want the i. in the cluster statement. While xi is not needed for recent Stata versions, I don't know about user-written ivreg2.

Dear Bromiley!

Thanks for your suggestions and help. It really works out when i delete the term "i." in the cluster statement, however, another issue confused me as i add one more iv, here's my codes:

xi: ivreg2 y3 (itnmarry = intmarry = iv_prefect0522 iv_05prob2) male age agesq i.educ i.employer2 i.industry i.marea i.home_provin i.p_prefect [pw = w_l], cluster(home_pref p_prefect) partial(i.home_provin i.p_prefect)

My question is how to obtain Hansen J test statistic for over identification test.

[Leo wong]

I worked it out by two ways: mata and chi2 test:
Here's my codes:
*.(1) 2SLS, IVs: iv_prefect0522 iv_05prob2

eststo clear

foreach var of varlist y* {
qui xi: ivreg2 `var' (intmarry = iv_prefect0522 iv_05prob2) `x1' i.educ i.employer2 i.industry i.marea `x2' i.s_educ `x3' i.home_provin i.p_prefect [pw = w_l], gmm2s cluster(home_pref p_prefect) partial(i.home_provin i.p_prefect)
eststo `var'_iv1
}
outreg2 [*_iv1] using iv1, keep(intmarry) excel adds(Weak identification test, `e(widstat)') dec(3) ct(`var') replace

*.(2) Calculate sargan statistic

preserve
gen cons = 1
predict res, resid
keep if e(sample)
keep iv_prefect0522 iv_05prob2 male age agesq tmperiod _Ieduc_1-_Ieduc_3 _Imarea_2 _Imarea_3 s_age _Is_educ_1-_Is_educ_3 h_dfct _Ihome_prov_12-_Ihome_prov_65 _Ip_prefect_2-_Ip_prefect_308 cons res
order iv_prefect0522 iv_05prob2 male age agesq tmperiod _Ieduc_1-_Ieduc_3 _Imarea_2 _Imarea_3 s_age _Is_educ_1-_Is_educ_3 h_dfct _Ihome_prov_12-_Ihome_prov_65 _Ip_prefect_2-_Ip_prefect_308 cons res

clear mata
mata:
data = st_data(.,.)
a = cols(data) - 1
X = st_data(.,1..a)
e = st_data(.,"res")
V = e' * X * invsym(X' * X) * X' * e
M = e' * e
st_numscalar("V",V)
st_numscalar("M",M)
end

scalar sargan = (e(N) * V) / M
di %4.2f sargan
di %4.3f chi2tail(1,sargan)
restore

*.(3) Follow the wooldrege

use wjw17, clear

local x1 male age agesq tmperiod
local x2 s_age
local x3 h_dfct

eststo clear
foreach var of varlist y* {
qui {
xi: ivreg2 `var' (intmarry = iv_prefect0522 iv_05prob3) `x1' i.educ i.marea `x2' i.s_educ `x3' i.home_provin i.p_prefect [pw = w_l], cluster(home_pref p_prefect) partial(i.home_provin i.p_prefect)
eststo `var'_iv1
preserve
predict res,resid
xi: reg res iv_prefect0522 iv_05prob2 `x1' i.educ i.marea `x2' i.s_educ `x3' i.home_provin i.p_prefect, r
scalar sargan=e(N)*e(r2)
}
di %4.2f sargan
di %4.3f chi2tail(1,sargan)
restore
}