Displaying 1 result from an estimated 1 matches for "qpglm".
Did you mean:
pglm
2006 Jul 21
2
glm cannot find valid starting values
...nterval t. The data provided are a subset of the full data set.
#M is the cumulative sum of individuals that have changed state up to t-1.
#Assume that the rate of state change is constant (S ~ kM), but the
#distribution of S is clustered.
#The goal is to estimate N.
#N can be estimated by fitting:
qpglm <- glm(S ~ M, family = quasipoisson(link = identity), sdat)
summary(qpglm)
N.est <- -coef(qpglm)[1]/coef(qpglm)[2]
N.est
#i.e. x-intercept is minus intercept/slope
#To estimate confidence limits on N.est, fit models without intercept to
#N.est - M + x, where x is an integer. The model will ha...