<>
Edwin wrote
i have a view X on k variables and a (k x k) matrix W
i need a vector where row i equals X[i,.]*W*X[i,.]'
at the moment i do the following
xwx = J(st_nobs(), 1, .)
for (i=1; i<=st_nobs(); ++i)
xwx[i] = X[i,.]*W*X[i,.]'
which does the trick, but i was wondering whether it is possible to
speed
this up?
This brings about a 20% speed improvement:
webuse auto,clear
drop make
drop if mi(rep78)
expand 10000
set rmsg on
mata:
st_view(X=.,.,.)
W = runiform(11,11)
end
mata:
xwx = J(st_nobs(), 1, .)
for (i=1; i<=st_nobs(); i++) {
xwx[i] = X[i,.]*W*X[i,.]'
}
end
mata:
XW = X*W
xwx2 = J(st_nobs(), 1, .)
for (i=1; i<=st_nobs(); i++) {
xwx2[i] = XW[i,.] * X[i,.]'
}
end
Kit
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
