Fitting Log-Linear Models by Message Passing
Fit log-linear models to multidimensional contingency tables by Iterative Proportional Fitting.
effloglin(table, margin, fit = FALSE, eps = 0.01, iter = 20, print = TRUE)
table |
A contingency table |
margin |
A generating class for a hierarchical log–linear model |
fit |
If TRUE, the fitted values are returned. |
eps |
Convergence limit; see 'details' below. |
iter |
Maximum number of iterations allowed |
print |
If TRUE, iteration details are printed. |
The function differs from loglin in that 1) data
can be given in the form of a list of sufficient marginals and
2) the model is fitted only on the cliques of the triangulated
interaction graph of the model. This means that the full table
is not fitted, which means that effloglin is efficient
(in terms of storage requirements). However effloglin is
implemented entirely in R and is therefore slower than
loglin. Argument names are chosen so as to match those
of loglin()
A list.
Søren Højsgaard, sorenh@math.aau.dk
Radim Jirousek and Stanislav Preucil (1995). On the effective implementation of the iterative proportional fitting procedure. Computational Statistics & Data Analysis Volume 19, Issue 2, February 1995, Pages 177-189
data(reinis)
glist <-list(c("smoke", "mental"), c("mental", "phys"),
c("phys", "systol"), c("systol", "smoke"))
stab <- lapply(glist, function(gg) tabMarg(reinis, gg))
fv3 <- effloglin(stab, glist, print=FALSE)Please choose more modern alternatives, such as Google Chrome or Mozilla Firefox.