Regression Analysis for Current Status Data - Interval-Censored-Time-to-Event Data: Methods and Applications - page 106

the animals in the conventional environment had higher tumor incidence rates

than those in the germ-free environment.
Now we consider the data introduced in Dunson and Dinse (2002). This

was a 2-year tumorigenicity study conducted by the National Toxicology Pro-

gram on groups of 50 male and female F344/N rats and B6C3F 1 mice. In the
study, the animals were exposed to chloroprene at different concentrations by

inhalation, 6 hours per day, 5 days per week, for 2 years with the goal of ex-

amining the effect of chloroprene on tumor growth. At their death or sacrifice,

the animals were examined for tumor presence or absence. For the illustration

here, we focus on two lung tumors|Alveolar/Bronchiolar Carcinoma (A/B

C) and Alveolar/Bronchiolar Adenoma (A/B A) |and consider the 100 male

B6C3F 1 mice in the control and high-dose groups. For the analysis, define T1i T1i
and T T2i to be the occurrence times of A/B C and A/B A, respectively, for
the i-th animal and Zi i = 1 if the i-th animal was in the high-dose group and
Z i = 0 otherwise, i = 1;:::; 100. Here, C i is the death or sacrifice time for
the i-th animal.
Assume that T T1i and T T2i follow the models in Equations (4.4) through (4.6).
Then the application of the estimation procedure with covariate-dependent

^ D = 2:3259 and ^ D = 13:3645, with the es-

monitoring times yielded

timated standard errors being 0.2747 and 19.1911, respectively. These results

indicate that the association parameter does not seem to depend on the treat-

ment. Next we applied the estimation procedure with covariate-independent
monitoring times obtained ^ = 2:3188 and ^ = 0:6085 with the estimated

standard errors of 0.1990 and 0.1666. Here for both cases, we assumed the

Clayton model for the joint distribution of T2i T1i and T T2i and used the same
kernel function and bandwidth as those used in the simulation studies. The

results suggest that the high-dose chloroprene had a significant effect of in-

creasing the tumor occurrence rates, and the occurrence rates of the two lung

tumors, A/B C and A/B A, were significantly negatively correlated.

To investigate the possible dependence of the analysis results given above