missmodel
object produced by an EM algorithm.
Loglin
,
Gauss
and
cgm
.
worstFraction(object, ...)
"missmodel"
object produced by an EM algorithm.
method
,
tol
,
lim
, and
h
which affect the algorithm used to find the worst fraction.
Additional arguments are possible. See the specific function called
for a list of all possible arguments.
fraction
component
estimates the largest eigenvalue of the
jacobian of the EM operator.
direction
component contains the corresponding eigenvector.
The direction is normalized to have length
one.
For an EM algorithm, convergence is approximately linear, with rate governed by the largest eigenvalue of the jacobian of the EM operator.
C. Fraley, "On Computing the Largest Fraction of Missing
Information for the EM Algorithm and the Worst Linear
Function for Data Augmentation", to appear in Computational Statistics
and Data Analysis.
Schafer, J. L. (1997), Analysis of Incomplete Multivariate Data,
Chapman & Hall, London.
fit <- emLoglin(crime, frequency= count) worstFraction(fit) worstFraction.Loglin(fit) #same