The identifiability of parameters may be examined by refitting a model with cloned data (each capture history replicated nclone times). For identifiable parameters the estimated variances are proportional to 1/nclone.
cloned.fit(object, nclone = 100, newdata = NULL, linkscale = FALSE)
Dataframe with columns* --
original estimate
original SE
cloned estimate (xxx = nclone)
cloned SE
SE.estimate / SE.estimate.xxx / sqrt(nclone)
* `estimate' becomes `beta' when linkscale = TRUE.
previously fitted openCR object
integer number of times to replicate each capture history
optional dataframe of values at which to evaluate model
logical; if TRUE then comparison uses SE of linear predictors
The key output is the ratio of SE for estimates from the uncloned and cloned datasets, adjusted for the level of cloning (nclone). For identifiable parameters the ratio is expected to be 1.0.
Cloning is not implemented for spatial models.
The comparison may be done either on the untransformed scale (using approximate SE) or on the link scale.
Lele, S.R., Nadeem, K. and Schmuland, B. (2010) Estimability and likelihood inference for generalized linear mixed models using data cloning. Journal of the American Statistical Association 105, 1617--1625.
openCR.fit
if (FALSE) {
fit <- openCR.fit(dipperCH)
cloned.fit(fit)
}
Run the code above in your browser using DataLab