Page 291 - Contributed Paper Session (CPS) - Volume 7

CPS2102 Iris Reinhard
4. Discussion and Conclusion
In this study a two-part mixed model for semicontinuous longitudinal data
is considered and evaluated by simulation. The model performs well for the
simulated data which represent the structure of real data from health
economy. The estimated effects are very close to their theoretical values. With
increasing sample size the MSE decreases and the α-level is maintained from
N≈100. Using the classical linear mixed model has to be discouraged because
it produces inflated type I error and much higher MSEs for the maximum
likelihood parameters, in particular when the effect is located only in the zero
component. The estimation can be implemented in SAS using the procedure
NLMIXED. A major advantage is the simultaneous modelling and estimation
of the two components. An alternative distribution to model the non-zero
values can be the lognormal or the gamma distribution for the continuous
part. Two-part mixed models encompass a considerable subset of parametric
models for semicontinuous longitudinal data previously suggested in the
literature.

278 | I S I W S C 2 0 1 9