-
Notifications
You must be signed in to change notification settings - Fork 1
/
MLEfunctions.R
95 lines (78 loc) · 3.04 KB
/
MLEfunctions.R
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
dbetabinom <- function (y, size, p, d){
a <- d * p
b <- d * (1 - p)
exp(lbeta(y + a, size - y + b) - lbeta(a, b) + lchoose(size, y))
}
getProbsMaster <- function(pCom, dCom, pHouse, dHouse, maxSize){
if(dCom == 0){
m <- matrix(0, maxSize+1, maxSize, dimnames = list(0:maxSize, 1:maxSize))
m[1,1:maxSize] <- 1-pCom
m[row(m)-col(m) == 1] <- pCom
}else if(dCom == Inf & dHouse == Inf){
m <- matrix(0, maxSize+1, maxSize+1, dimnames = list(0:maxSize, 0:maxSize))
m[1,1] <- 1
for(k in 1:maxSize){
m[1:k,k+1] <- choose(k,0:(k-1)) * diag(m)[1:k] * (1-pCom)^(k:1) * (1-pHouse)^((0:(k-1))*(k:1))
m[k+1,k+1] <- 1 - sum(m[,k+1])
}
m <- m[,-1]
}else if(dHouse == 0){
m <- matrix(0, maxSize+1, maxSize, dimnames = list(0:maxSize, 1:maxSize))
noTransm <- (1-pHouse)^(0:(maxSize-1))
k <- sequence(1:maxSize)
n <- rep(1:maxSize, 1:maxSize)
if(dCom == Inf) db <- dbinom(k-1,n,pCom) else db <- dbetabinom(k-1,n,pCom,dCom)
m[upper.tri(m,diag=TRUE)] <- noTransm[k] * db
m[row(m)-col(m) == 1] <- 1 - colSums(m)
}else{
m <- matrix(0, maxSize+1, maxSize, dimnames = list(0:maxSize, 1:maxSize))
if(dCom == Inf){
m[1,1:maxSize] <- (1-pCom)^(1:maxSize)
dbCom <- dbinom(sequence(1:maxSize),rep(1:maxSize,1:maxSize),pCom)
}else{
m[1,1:maxSize] <- dbetabinom(0,1:maxSize,pCom,dCom)
dbCom <- dbetabinom(sequence(1:maxSize),rep(1:maxSize,1:maxSize),pCom,dCom)
}
m[2,1] <- 1 - m[1,1]
pc <- matrix(0,maxSize,maxSize)
pc[upper.tri(pc,diag=TRUE)] <- dbCom
if(pHouse == 0){
m[2:(maxSize+1),1:maxSize] <- pc
}else{
M <- maxSize-1
M2 <- M^2
ph <- array(0,c(M,M,M))
if(dHouse == Inf){ for(j in 0:(M-1)) ph[1,j+1,(j+1):M] <- dbinom(j,(j+1):M,pHouse)
}else for(j in 0:(M-1)) ph[1,j+1,(j+1):M] <- dbetabinom(j,(j+1):M,pHouse,dHouse)
for(i in 2:M)
for(j in 0:(M-i))
for(k in (j+1):(M-i+1))
ph[i,j+1,k] <- sum(ph[i-1,1:(j+1),k] * ph[1 + M*(j:0) + M2*((k-1):(k-j-1))])
pt <- ph
for(j in 1:(M-1))
for(i in 1:(M-j))
for(k in (j+1):(M-i+1))
pt[i,j+1,k] <- sum(ph[i,(j+1):2,k] * pt[j:1 + (0:(j-1))*M + ((k-j-1):(k-2))*M2])
for(n in 2:maxSize){
for(k in 1:(n-1)) m[k+1,n] <- sum(pc[k:1,n] * pt[k:1 + (0:(k-1))*M + ((n-k-1):(n-2))*M2])
m[n+1,n] <- 1 - sum(m[,n])
}
}
}
m
}
getP <- function(pCom,dCom,pHouse,dHouse,phiV,phiA,piV,piA){
p <- 0
if(phiA > 0 & phiA <= 1 & phiV > 0 & phiV <= 1 & pCom > 0 & pCom < 1 & pHouse >= 0 & pHouse < 1 & dHouse >= 0 & dCom >= 0 &
piV <= 1 & piA <= 1 & piV > 0 & piA > 0){
M <- getProbsMaster(pCom,dCom,pHouse,dHouse,maxSize)
x <- M[Mi] * coef * phiV^tpV * (1-phiV)^fnV * phiA^tpA * (1-phiA)^fnA *
piV^tnV * (1-piV)^fpV * piA^tnA * (1-piA)^fpA
p <- tapply(x,rep(seq_along(numTerms),numTerms),sum)
}
p
}
getLL <- function(pCom,dCom,pHouse,dHouse,phiV,phiA,piV,piA)
sum(cf$freq * log(getP(pCom,dCom,pHouse,dHouse,phiV,phiA,piV,piA)))
getSol <- function(initpar,fn)
optim(initpar,fn,control=list(abstol = 1e-10, maxit=10000))