SIMEX algorithm

Therefore, the SIMEX algorithm is as follows. First a simulation step is performed ... 

In order to use the SIMEX algorithm on this data, an estimate of the measurement variance for creatinine clearance must be obtained. In discussions with clinical chemists, the upper limit of measurement error associated with measuring serum or urine creatinine using the Jaffe Reaction is 5%. Assuming mean 24 hour values for urinary volume, urinary daily creatinine excretion, and serum creatinine of 1000 mL, 1.5 g, and 1.1 mg/dL, respectively, an approximate measurement error variance for creatinine clearance was found to be 60 (mL/min)2. 

With this as an estimate of the assay measurement variance, the SIMEX algorithm was applied. Figure 2.10 plots the mean regression parameter against varying values of X using 1000 iterations for each value of X. Extrapolation of X to — 1 for both the slope and intercept leads to a SIMEX equation of... 

The same caveats that apply to linear models when the predictor variables are measured with error apply to nonlinear models. When the predictor variables are measured with error, the parameter estimates may become biased, depending on the nonlinear model. Simulation may be used as a quick test to examine the dependency of parameter estimates within a particular model on measurement error (Fig. 3.14). The SIMEX algorithm, as introduced in the chapter on Linear Models and Regression, can easily be extended to nonlinear models, although the computation time will increase by orders of magnitude. 

To further understand the impact of measurement error on the model parameters, the SIMEX algorithm was applied using a model containing only the 1,5-h and 4-h time points. The assay had a coefficient of variation of about 5% based on the bioanalytical validation of standards and controls in plasma having a known amount of drug added. The resulting SIMEX model was... 

---