# Evaluation of Graphical Output

Subjective evaluation of graphical output such as the calculated and observed data versus time plot and the weighted residual plots can be very useful in the determination of the best model.

## Calculated and Observed Data versus Time Plots

Look for systematic deviations. Could the deviations be explained with an addition to the model? Is a distribution phase obvious. Are there straight lines portion on semi-log plots, linear plots? Combining data from different sample sites, such as plasma and urine data, in one analysis will increase the model size and complexity. Figure 19.2.1 Linear plot of Calculated and Observed Data versus Time

Note the systemic, significant deviations between the line labeled 'Calculated (1c)'. This line represents the 'best' fit to a one compartment model. The two compartment (2c) model represents a much better fit to the observed data.

## Weighted Residual Plots

Another plot that can be very useful in determining the 'best' model is the weighted residual plot Figure 19.2.2 Linear plot of Weighted Residual versus Time

A pattern in the weighted residual plot may indicate that the model used is too small. In the case of Figure 19.2.2 there is a definite 'U' or 'V' shape suggesting that a bigger model (with two more parameters) should be investigated. A diagonal line would suggest a model with one additional adjustable parameter could be useful.