forked from ianfiske/unmarked
-
Notifications
You must be signed in to change notification settings - Fork 25
/
tmb_gdistremoval.hpp
203 lines (165 loc) · 5.46 KB
/
tmb_gdistremoval.hpp
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
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
#undef TMB_OBJECTIVE_PTR
#define TMB_OBJECTIVE_PTR obj
// name of function below **MUST** match filename
template <class Type>
Type tmb_gdistremoval(objective_function<Type>* obj) {
//Describe input data
DATA_VECTOR(y_dist); //observations
DATA_VECTOR(y_rem);
DATA_MATRIX(y_sum);
DATA_INTEGER(mixture);
DATA_INTEGER(K);
DATA_IVECTOR(Kmin);
DATA_INTEGER(T);
DATA_MATRIX(X_lambda); //lambda fixed effect design mat
DATA_SPARSE_MATRIX(Z_lambda); //psi random effect design mat
DATA_INTEGER(n_group_vars_lambda); //# of grouping variables for lambda
DATA_IVECTOR(n_grouplevels_lambda); //# of levels of each grouping variable
DATA_MATRIX(X_phi);
DATA_SPARSE_MATRIX(Z_phi);
DATA_INTEGER(n_group_vars_phi);
DATA_IVECTOR(n_grouplevels_phi);
DATA_MATRIX(X_dist);
DATA_SPARSE_MATRIX(Z_dist);
DATA_INTEGER(n_group_vars_dist);
DATA_IVECTOR(n_grouplevels_dist);
DATA_MATRIX(X_rem);
DATA_SPARSE_MATRIX(Z_rem);
DATA_INTEGER(n_group_vars_rem);
DATA_IVECTOR(n_grouplevels_rem);
DATA_INTEGER(keyfun_type);
DATA_VECTOR(A); // Area
DATA_VECTOR(db); // distance breaks
DATA_MATRIX(a);
DATA_VECTOR(w);
DATA_MATRIX(u);
DATA_IVECTOR(per_len); // Length of removal periods
PARAMETER_VECTOR(beta_lambda); //Fixed effect params for lambda
PARAMETER_VECTOR(b_lambda); //Random intercepts and/or slopes for lambda
PARAMETER_VECTOR(lsigma_lambda); //Random effect variance(s) for lambda
PARAMETER_VECTOR(beta_alpha); //Only used if NB or ZIP
Type log_alpha = 0;
if(mixture > 1) log_alpha = beta_alpha(0);
PARAMETER_VECTOR(beta_phi);
PARAMETER_VECTOR(b_phi);
PARAMETER_VECTOR(lsigma_phi);
PARAMETER_VECTOR(beta_dist); //Same thing but for det
PARAMETER_VECTOR(b_dist);
PARAMETER_VECTOR(lsigma_dist);
PARAMETER_VECTOR(beta_scale); //Trick here: this is 0-length array if keyfun != hazard
Type scale = 0; // If not hazard this is ignored later
if(keyfun_type == 3) scale = exp(beta_scale(0)); // If hazard
PARAMETER_VECTOR(beta_rem); //Same thing but for det
PARAMETER_VECTOR(b_rem);
PARAMETER_VECTOR(lsigma_rem);
Type loglik = 0.0;
int M = X_lambda.rows(); // # of sites
int Rdist = y_dist.size() / M;
int Jdist = Rdist / T;
int Rrem = y_rem.size() / M;
int Jrem = Rrem / T;
//Construct lambda vector
vector<Type> lam = X_lambda * beta_lambda;
lam = add_ranef(lam, loglik, b_lambda, Z_lambda, lsigma_lambda,
n_group_vars_lambda, n_grouplevels_lambda);
lam = exp(lam);
lam = lam.array() * A.array();
//Construct availability (phi) vector
vector<Type> phi(M*T);
phi.setOnes();
if(T > 1){
phi = X_phi * beta_phi;
phi = add_ranef(phi, loglik, b_phi, Z_phi, lsigma_phi,
n_group_vars_phi, n_grouplevels_phi);
phi = invlogit(phi);
}
//Construct distance parameter (sigma, rate, etc.) vector
vector<Type> dp(M*T);
if(keyfun_type > 0){ // If keyfun is not uniform
dp = X_dist * beta_dist;
dp = add_ranef(dp, loglik, b_dist, Z_dist, lsigma_dist,
n_group_vars_dist, n_grouplevels_dist);
dp = exp(dp);
}
//Construct removal parameter vector
vector<Type> rp(M*Rrem);
rp = X_rem * beta_rem;
rp = add_ranef(rp, loglik, b_rem, Z_rem, lsigma_rem,
n_group_vars_rem, n_grouplevels_rem);
rp = invlogit(rp);
//Likelihood
for (int i=0; i<M; i++){
Type site_lp = 0;
// Calculate 2nd abundance parameter
Type alpha = 0;
Type var;
if(mixture == 2){
alpha = exp(log_alpha);
var = lam(i) + pow(lam(i), 2) / alpha; // TMB parameterization
} else if(mixture == 3){
alpha = invlogit(alpha);
}
vector<Type> f(K+1);
f.setZero();
// Iterate over possible true abundance values
for (int k=Kmin(i); k<(K+1); k++){
if(mixture == 2){
f(k) = dnbinom2(Type(k), lam(i), var, false);
} else if(mixture == 3){
f(k) = dzipois(Type(k), lam(i), alpha, false);
} else {
f(k) = dpois(Type(k), lam(i), false);
}
}
int t_ind = i * T;
int yd_ind = i * T * Jdist;
int yr_ind = i * T * Jrem;
vector<Type> yd_sub(Jdist);
vector<Type> yr_sub(Jrem);
vector<Type> rp_sub(Jrem);
vector<Type> cpd(Jdist);
vector<Type> cpr(Jrem);
Type pdist;
Type prem;
Type pall;
vector<Type> g(K+1);
g.setOnes();
vector<Type> fg(K+1);
vector<Type> asub = a.row(i);
vector<Type> usub = u.row(i);
for (int t=0; t<T; t++){
yd_sub = y_dist.segment(yd_ind, Jdist);
yr_sub = y_rem.segment(yr_ind, Jrem);
rp_sub = rp.segment(yr_ind, Jrem);
//If any NAs found skip site
for (int j=0; j<Jdist; j++){
if(R_IsNA(asDouble(yd_sub(j)))) goto endper;
}
for (int j=0; j<Jrem; j++){
if(R_IsNA(asDouble(yr_sub(j)))) goto endper;
}
cpd = distance_prob(keyfun_type, dp(t_ind), scale, 1, db, w, asub, usub);
pdist = sum(cpd);
cpd = cpd/pdist;
site_lp += dmultinom(yd_sub, cpd, true);
cpr = pifun_removal(rp_sub, per_len);
prem = sum(cpr);
cpr = cpr/sum(cpr);
site_lp += dmultinom(yr_sub, cpr, true);
pall = pdist * prem * phi(t_ind);
for (int k=Kmin(i); k<(K+1); k++){
g(k) *= dbinom(y_sum(i,t), Type(k), pall, false);
}
endper: ;
t_ind += 1;
yd_ind += Jdist;
yr_ind += Jrem;
}
fg = f.array() * g.array();
site_lp += log(sum(fg));
loglik -= site_lp;
}
return loglik;
}
#undef TMB_OBJECTIVE_PTR
#define TMB_OBJECTIVE_PTR this