Biomedical Engineering Reference
In-Depth Information
to represent a treatment group. Options
EMstep
and
ICMstep
specify whether
an EM, ICM, or the hybrid EM-ICM algorithm (Wellner and Zhan, 1997) will
be used to estimate the common survival function when assuming that all
survival functions are identical. By default, the EM-ICM hybrid approach is
taken. The other two inputs
tol
and
maxiter
specify the convergence criterion
and upper bound for the number of iterations allowed in the estimation of the
common survival function. Note that an EM estimate may not be the MLE
but an EM-ICM estimate always is.
In
addition,
to
conduct
a
generalized
logrank
test,
the
function
ModifiedEMICM
in glrt computes an EM estimate or MLE of the survival
function based on the first two columns of the data.
Package glrt includes a set of interval-censored data named
diabetes
based on a diabetes study. More details about the data and study are intro-
duced in Section 14.4.
14.4
Applications
14.4.1
Conducting Generalized Logrank Tests
Breast Cosmesis Study: The study involved 94 early breast cancer pa-
tients. The objective of the study was to compare the patients who had been
treated with radiotherapy alone (treatment 1, 46 patients) to those treated
with primary radiation therapy and adjuvant chemotherapy (treatment 2, 48
patients). The survival time of interest was the time until the appearance of
breast retraction. In the study, the patients were monitored for breast retrac-
tion every 4 to 6 months. However, they often missed visits as their recovery
progressed and returned in a changed status. Thus, only interval-censored data
on the survival time were observed. The data first appeared in Finkelstein and
Wolfe (1985) and have been analyzed by many researchers. The package Icens
Search WWH ::
Custom Search