Displaying 3 results from an estimated 3 matches for "acwrk".
2010 Sep 07
5
question on "optim"
...ccl<-matrix(1,n,ns)
eta<-c(0,0)
# setup the parts for loglikelihood
q1<-exp(par[1])
pr1<-q1/(1+q1)
pr2<-1-pr1
eta[2]<-par[2]
a<-par[3:6]
b<-par[7:11]
w<-par[12:npar]
for(m in 1:ns){
?for(i in 1:nt){
?? regw[,i]<-w[1]+ w[2]*eta[m]+exp(w[3]+w[4]*eta[m])*actr[,i]+w[5]*acwrk[,i]
??? vbar2[,i]=a[1]+???? eta[m]+regw[,i]*a[2]+acwrk[,i]*a[3]+actr[,i]*a[4]
??? vbar3[,i]=b[1]+b[2]*eta[m]+regw[,i]*b[3]+acwrk[,i]*b[4]+actr[,i]*b[5]
??? v8[,i]=1+exp(vbar2[,i])+exp(vbar3[,i])
????
?? for(j in 1:n){
?? ?if (home[j,i]==1) lia[j,i]=1/v8[j,i]
????? if (wrk[j,i]==1) lia[j,i]=exp(vba...
2010 Aug 17
1
replace loops with matrix
...0 of the time the loop
method takes. it may still have room to improve it. at least, the?following
part?looks ugly.
ccl[,m]<-lia[,1]*lia[,2]*lia[,3]*lia[,4]*lia[,5]
any suggestion are appreciated.
The Loop code:
for(m in 1:ns){
?for(i in 1:nt){
??? vbar2[,i]=a[1]+???? eta[m]+acedu[,i]*a[2]+acwrk[,i]*a[3]
??? vbar3[,i]=b[1]+b[2]*eta[m]+acedu[,i]*b[3]+acwrk[,i]*b[4]
??? v8[,i]=1+exp(vbar2[,i])+exp(vbar3[,i])
????? for(j in 1:n){
?? ???? ? if (edu[j,i]==1) lia[j,i]=1/v8[j,i]
????????? if (wrk[j,i]==1) lia[j,i]=exp(vbar2[j,i])/v8[j,i]
????????? if (home[j,i]==1) lia[j,i]=exp(vbar3[j,i])/v8[j...
2010 Sep 15
1
optim with BFGS--what may lead to this, a strange thing happened
...lia<-matrix(0,n,nt)
ccl<-matrix(1,n,ns)
eta<-c(0,0)
# setup the parts for loglikelihood
q1<-exp(par[1])
pr1<-q1/(1+q1)
pr2<-1-pr1
eta[2]<-par[2]
a<-par[3:6]
b<-par[7:11]
w<-par[12:npar]
for(m in 1:ns){
regw<-exp(w[1]+w[2]*eta[m])*actr+exp(w[3]+ w[4]*eta[m])*acwrk
vbar2=a[1]+ eta[m]+regw*a[2]+acwrk*a[3]+actr*a[4]
vbar3=b[1]+b[2]*eta[m]+regw*b[3]+acwrk*b[4]+actr*b[5]
v8=1+exp(vbar2)+exp(vbar3)
lia<-ifelse(home==1,1/v8,
ifelse(wrk==1,exp(vbar2)/v8,
ifelse(tr==1,exp(vbar3)/v8,1)))
wden<-ifelse(wrk==1,...