Researchers are interested in the rate of declining CD4+ cell count after stopping antiretroviral therapy.  For example, patients who 
entered the "Strategies for Management of Anti-retroviral Therapy"(SMART) study while on antiretroviral therapy, and who got 
randomized to the DC arm, were stopping antiretrovirals, and we investigate statistical models that could be used to estimate their rate 
of CD4+ decline within the first two months. 

It is well-known that the CD4+ cell count at study entry (baseline) is associated with the rate of CD4+ decline, yet the "true" baseline 
CD4+ is unknown, since it is measured with an error. In this paper we investigate the influence of model misspecification on the estimates 
of the slope parameters; the misspecification occurs because we use the "observed" baseline CD4+ to fit data that were simulated with an 
association between slope and "true" baseline CD4+.  We present simulation studies and compare the results from misspecified models 
with true underlying models.  Then we apply our misspecified models to a subset of the SMART data.