Estimation of std-error for canonical correlation (example in STATA)

Hi all,

I'm puzzled by one output of STATA. It is shown on this web-site (look
for 'Standard errors estimated conditionally')
http://www.ats.ucla.edu/stat/stata/examples/cama3/cama3stata_ch10.htm

In the third column, there's an estimate of the standard error. Does
anybody know how to they do it? I was thinking about bootstrapping, but
I'm not sure that it is what they use.

Thanks in advance,
Tony

.