Page 1 of 1

near singular matrix

Posted: Sun Mar 23, 2014 1:00 pm
by maoqi
I want to ask:
there are 3 transformations of my system, why the first function system reflected by eviews as "near singular matrix", but the other two which are similar can be calculated by eviews. Please tell me some clues. Thank you very much.
1st:

log(es)-log(es(-1)) - log(y) +log(y(-1)) = c(1)*(c(2) - 1) + c(2)*( log(py) - log (py(-1)) - log(pe) + log(pe(-1)))

log(cs/(cs+ls)) - log(cs(-1)/(cs(-1) + ls(-1))) = c(3)*(c(4) - 1) + (c(4) - 1)/(1-c(2))*(log((cs+ls)/1) - log((cs(-1)+ls(-1))/1)) + (1- c (4))*(log(pk) - log(pk(-1)) - log(py)+log(py(-1)))

log(ls/(cs+ls)) - log(ls(-1)/(cs(-1) + ls(-1))) = c(5)*(c(4) - 1) + (c(4) - 1)/(1-c(2))*(log((cs+ls)/1) - log((cs(-1)+ls(-1))/1)) + (1- c (4))*(log(pl) - log(pl(-1)) -log(py)+log(py(-1)))


2nd and 3rd are calculated by eviews:

2nd:
log(cs)-log(cs(-1)) - log(y) +log(y(-1)) = c(6)*(c(7) - 1) + c(7)*( log(py) - log (py(-1)) - log(pk) + log(pk(-1)))

log(es/(es+ls)) - log(es(-1)/(es(-1) + ls(-1))) = c(8)*(c(9) - 1) + (c(9) - 1)/(1-c(7))*(log((es+ls)/1) - log((es(-1)+ls(-1))/1)) + (1- c (9))*(log(pe) - log(pe(-1)) - log(py)+log(py(-1)))

log(ls/(es+ls)) - log(ls(-1)/(es(-1) + ls(-1))) = c(10)*(c(9) - 1) + (c(9) - 1)/(1-c(7))*(log((es+ls)/1) - log((es(-1)+ls(-1))/1)) + (1- c (9))*(log(pl) - log(pl(-1)) -log(py)+log(py(-1)))


3rd:
log(ls)-log(ls(-1)) - log(y) +log(y(-1)) = c(11)*(c(12) - 1) + c(12)*( log(py) - log (py(-1)) - log(pl) + log(pl(-1)))

log(cs/(cs+es)) - log(cs(-1)/(cs(-1) + es(-1))) = c(13)*(c(14) - 1) + (c(14) - 1)/(1-c(12))*(log((cs+es)/1) - log((cs(-1)+es(-1))/1)) + (1- c (14))*(log(pk) - log(pk(-1)) - log(py)+log(py(-1)))

log(es/(cs+es)) - log(es(-1)/(cs(-1) + es(-1))) = c(15)*(c(14) - 1) + (c(14) - 1)/(1-c(12))*(log((cs+es)/1) - log((cs(-1)+es(-1))/1)) + (1- c (14))*(log(pe) - log(pe(-1)) -log(py)+log(py(-1)))

Re: near singular matrix

Posted: Sun Mar 23, 2014 1:55 pm
by NicolasR
Did you tried different starting values for the coefficients? Maybe the coefficients c(1) to c(5) have a value where the algorithm can not start its iteration.

You can change them in the c vector coeficient.