Calculating the inverse of a variance covariance matrix in stata error 3251

[Donald Merkuri]

Hello statalist members i have the variance covariance matrix of BG that is a matrix that combines two vectors created as follows:
mat BG=G,B
mat li BG
// var-covar matrix
mat var_cov=get(VCE)
mat li var_cov

I have a problem when i try to calculate the inverse of var_cov matrix since when i input:

mata: inv_cov=cholinv("var_cov")

it pops out error 3251: cholinv(): 3251 <tmp> [1,1] found where real or complex required
<istmt>: - function returned error

My values should be all real and i don't understand why it tells me that they are not.
Anybody has an answer?
Would really appreciate your help

Best
Donald

[JanDitzen]

Hi Donald,
the code should work if you write:

mata: inv_cov=cholinv(var_cov)

So just leave out the quotes.
Best.
Jan

[Christophe Kolodziejczyk]

var_cov is not a mata object, but a Stata matrix

mata:
v = st_matrix("var_cov")
inv_cov=cholinv(v)

end

Or

mata: inv_cov=cholinv(st_matrix("var_cov"))

[Christophe Kolodziejczyk]

If you work in interactive mode and you want to copy the content of a Stata variable in the memory of Mata then you can either use getmata or st_data()/st_view(). In non-interactive mode you can only use st_data()/st_view().
If you want to get a matrix or a vector (different from a variable) then use st_matrix(). If you need a Stata scalar use st_numscalar(). If you need a local then use st_local(), etc.

More generally take a look at the "Stata" entry in the Mata manual. It is a good investment if you plan to use Mata in the future. The general principle is to transform Stata object into Mata objects. the st_* functions are here to just achieve that, i.e to allow Stata and Mata to communicate together.

Here is an example

clear
set obs 10

drawnorm x1 x2

mat bStata = (1,1)

mata:
xMata = st_data(.,"x1 x2")
bMata = st_matrix("bStata")

product = x*b'

product


end