Check Whether One or More Parameter Combinations in a gnm Model are Identified
For each of a specified set of linear combinations of parameters from a
gnm model, checks numerically whether the combination's
estimate is invariant to re-parameterization of the model.
checkEstimable(model, combMatrix = diag(length(coef(model))),
tolerance = NULL)model |
a model object of class |
combMatrix |
numeric: either a vector of length the same as
|
tolerance |
numeric: a threshold value for detection of
non-estimability. If |
A logical vector of length equal to the number of parameter combinations
tested; NA where a parameter combination is identically zero.
David Firth and Heather Turner
Catchpole, E.A. and Morgan, B.J.T. (1997). Detecting parameter redundancy. Biometrika, 84, 187–196.
set.seed(1)
## Fit the "UNIDIFF" mobility model across education levels
unidiff <- gnm(Freq ~ educ*orig + educ*dest +
Mult(Exp(educ), orig:dest), family = poisson,
data = yaish, subset = (dest != 7))
## Check whether multiplier contrast educ4 - educ5 is estimable
ofInterest(unidiff) <- pickCoef(unidiff, "[.]educ")
mycontrast <- numeric(length(coef(unidiff)))
mycontrast[ofInterest(unidiff)[4:5]] <- c(1, -1)
checkEstimable(unidiff, mycontrast)
## should be TRUE
## Check whether multiplier educ4 itself is estimable
mycontrast[ofInterest(unidiff)[5]] <- 0
checkEstimable(unidiff, mycontrast)
## should be FALSE -- only *differences* are identified herePlease choose more modern alternatives, such as Google Chrome or Mozilla Firefox.