.BG
.FN twoway
.TL
Analysis of Two-way Table
.CS
twoway(x, trim=.5, iter=6, eps, print=F)
.AG x
matrix to be analyzed.  Missing values (`NA's) are allowed.
.AG trim
optional trimming fraction for carrying out analysis.
Default value corresponds to using medians.  `trim=0' will cause
analysis by means, .25 by midmeans, etc.
.AG iter
maximum number of full (row and column) sweeps.
.AG eps
error tolerance.  If `eps' is given, the algorithm will
iterate until the maximum change in row or column effects is
< `eps'.  Default is to iterate until the specified number
of iterations or until converged to the accuracy of the
machine arithmetic.  It is not always possible to converge
to a unique answer.
.AG print
if `TRUE', the maximum change in row/column effects in
the last iteration is printed.
.RT
list with 4 components, `resid', `row', `col',
and `grand', such that
.br
`x[i,j]' equals `grand + row[i] + col[j] + resid[i,j]'
.br
.RC grand
overall location estimate of the data.
.RC row
vector of row effects.
.RC col
vector of column effects.
.RC resid
matrix of residuals from the fit.
.br
.SP
.SA
Function `plotfit' produces a graphical display of the
fit generated by `twoway'.
.EX
twoway(temperature,trim=.25) # analysis by midmeans
.KW robust
.KW array
.KW multivariate
.WR
