| capture {repeated} | R Documentation |
capture fits the Cormack capture-recapture model to n sample
periods. Set n to the appropriate value and type eval(setup).
n <- periods # number of periods
eval(setup)
This produces the following variables -
p[i]: logit capture probabilities,
pbd: constant capture probability,
d[i]: death parameters,
b[i]: birth parameters,
pw: prior weights.
Then set up a Poisson model for log linear models:
z <- glm(y~model, family=poisson, weights=pw)
and call the function, capture.
If there is constant effort, then all estimates are correct.
Otherwise, n[1], p[1], b[1], are correct only if
there is no birth in period 1. n[s], p[s], are correct
only if there is no death in the last period. phi[s-1] is
correct only if effort is constant in (s-1, s). b[s-1]
is correct only if n[s] and phi[s-1] both are.
capture(z, n)
z |
A Poisson generalized linear model object. |
n |
The number of repeated observations. |
capture returns a matrix containing the estimates.
J.K. Lindsey
y <- c(0,1,0,0,0,1,0,1,0,0,0,1,0,0,0,14,1,1,0,2,1,2,1,16,0,2,0,11,
2,13,10,0)
n <- 5
eval(setup)
# closed population
print(z0 <- glm(y~p1+p2+p3+p4+p5, family=poisson, weights=pw))
# deaths and emigration only
print(z1 <- update(z0, .~.+d1+d2+d3))
# immigration only
print(z2 <- update(z1, .~.-d1-d2-d3+b2+b3+b4))
# deaths, emigration, and immigration
print(z3 <- update(z2, .~.+d1+d2+d3))
# add trap dependence
print(z4 <- update(z3, .~.+i2+i3))
# constant capture probability over the three middle periods
print(z5 <- glm(y~p1+pbd+p5+d1+d2+d3+b2+b3+b4, family=poisson, weights=pw))
# print out estimates
capture(z5, n)