[本週專題]WinBUGS: Undefined Real Result?

是 否 +2 论坛币

k人 参与回答

经管之家送您一份

应届毕业生专属福利!

求职就业群

赵安豆老师微信：zhaoandou666

经管之家联合CDA

送您一个全额奖学金名额~ !

立即领取

感谢您参与论坛问题回答

经管之家送您两个论坛币！

+2 论坛币

Attched please review my code for latent class model. Until I had done model check-up, there are no errors. However, when I tried to "update" the sample, suddenly the trap message said "undefined real result" popped up. How can I deal with this problem? Thanks

1. model { # Marginal tabulations of Latent Diagnosis against Observed Items for (i in 1:n) { for (j in 1:K) { for (k in 1:2) {M1[j,k,i] <- equals(T[i],j)*equals(Y[i,1],k-1)}}} for (j in 1:K) {for (k in 1:2) {Tab1[j,k] <- sum(M1[j,k,1:n])}}
   
2. 
   
3. for (i in 1:n) { for (j in 1:K) { for (k in 1:2) {M2[j,k,i] <- equals(T[i],j)*equals(Y[i,2],k-1)}}} for (j in 1:K) {for (k in 1:2) {Tab2[j,k] <- sum(M2[j,k,1:n])}}
   
4. 
   
5. for (i in 1:n) { for (j in 1:K) { for (k in 1:2) {M3[j,k,i] <- equals(T[i],j)*equals(Y[i,3],k-1)}}} for (j in 1:K) {for (k in 1:2) {Tab3[j,k] <- sum(M3[j,k,1:n])}}
   
6. 
   
7. Bernoulli sampling
   
8. for (s in 1:n) {for (j in 1:M) {Y[s,j] ~ dbern(pi[T[s],j]) # new item data Z[s,j] ~ dbern(pi[T[s],j])} for (h in 1:2) { for (i in 1:2) { for (j in 1:2) { g[h,i,j,s] <- equals(Z[s,1]+1,h)*equals(Z[s,2]+1,i)*equals(Z[s,3]+1,j)}}}}
   
9. 
   
10. Implied aggregate table (G.new)
    
11. for (h in 1:2) { for (i in 1:2) { for (j in 1:2) {G[h,i,j] <- sum(g[h,i,j,])}}}
    
12. 
    
13. for (k in 1:M) {for (j in 1:K) {logit(pi[j,k]) <- theta[j,k]}}
    
14. 
    
15. Posterior memberships
    
16. for (i in 1:n) { for (j in 1:K) {post[j,i] <- equals(T[i],j)}}
    
17. 
    
18. Priors
    
19. for (k in 1:M) {theta[1,k] ~ dnorm(0,0.1) I(,theta[2,k]) theta[2,k] ~ dnorm(0,0.1) I(theta[1,k],)} eta.s[1] ~ dgamma(w1,1) I(,eta.s[2]) eta.s[2] ~ dgamma(w1,1) I(eta.s[1],) for (j in 1:K) {eta[j] <- eta.s[j]/sum(eta.s[])} for (i in 1:n) {T[i] ~ dcat(eta[])}}
    
20. 
    
21. B: Data list( n = 103, # number of patients observed K = 2, # number of latent classes M = 3 , # number of diagnostic tests w1 =1,
    
22. 
    
23. observed item data
    
24. Y=structure(.Data=c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,1,0,0,1,0,0,1,0,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1),.Dim=c(103,3)))
    
25. 
    
26. B: Inits list(T=c(1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2, 1,2,1,2,1,2,1,2,1,1,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,1,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1),eta.s=c(0.5,0.5),theta=structure(.Data=c(0,0,0,0,0,0),.Dim=c(2,3)),Z=structure(.Data=c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,1,0,0,1,0,0,1,0,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1),.Dim=c(103,3)))
    
27. 
    
28. list(T=c(1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2, 1,2,1,2,1,2,1,2,1,1,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,1,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1,2,1),eta.s=c(0.427, 0.573),theta=structure(.Data=c(-2.5, -3.3, -2.7, 1.7, 1.4, 1.72),.Dim=c(2,3)),Z=structure(.Data=c(0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0, 0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,1,0,0,1,0,0,1,0,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,0,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1, 1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,0,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1),.Dim=c(103,3)))

复制代码

扫码加我 拉你入群

请注明：姓名-公司-职位

以便审核进群资格，未注明则拒绝

分享0 收藏0 回帖

关键词：Undefined winbugs Defined DEFINE WINBUG undefined suddenly message against problem

Error messages in WinBugs are terrible to interpret. I believe that this message is indicating that the program is unable to perform the calculations of the model due to illogical or extreme values selected by the sampler.

The first thing I would do is calculate values wherever you use a function. For example calculate the logit transformation by hand instead of using the logit function. I have fixed this exact error by making this change once.

Secondly, I would carefully go over your priors to make sure that they don't allow parameter values that could cause an error in calculation. For example,
d~dnorm(0,1)
c <- log(d)
should cause a trap error because the prior allows for negative values which cannot be log transformed. Another example is if the prior allows for extremely positive values of a parameter that will later be exponentiated. I don't know what the limit is on this, but obviously exp(10000000) will cause an error.

I would try giving it the best initial values that you can come up with.

Provide slightly informed priors.

Make sure that the looping is as efficient as it can be. Marc Kery's book Introduction to Winbugs for Ecologist provides some tips on this (http://www.lce.esalq.usp.br/arquivos/aulas/2010/LCE5813/Introduction%20to%20WinBUGS%20for%20Ecologists.pdf)

Lastly, I would perform the analysis in JAGS. Subtle differences between bugs and jags means that some problems can be resolve by simply switching programs. Not very satisfying but may work.

I hope this helps you resolve your trap error and move forward!