Table 2.
The results of pharmacodynamic model building and selection*
| Active moiety | Model | Number of model parameters | Model description | – LL** | AIC |
|---|---|---|---|---|---|
| DHM only | (1) | 2 | Linear effect (no intercept) | 171.508 | 4.388 |
| (2) | 3 | Linear effect (with intercept) | 171.255 | 4.431 | |
| (3) | 4 | Emax model | 169.996 | 4.400 | |
| DHC only | (4) | 2 | Linear effect (no intercept) | 182.728 | 4.668 |
| (5) | 3 | Linear effect (with intercept) | 174.485 | 4.512 | |
| (6) | 4 | Emax model | 175.149 | 4.529 | |
| DHC and DHM | (7) | 3 | Linear effect (no intercept) | 171.537 | 4.438 |
| (8) | 4 | General Emax model (no synergism or antagonism) | 171.419 | 4.485 | |
| (9) | 5 | General Emax model (with synergism or antagonism) | 171.850 | 4.546 | |
| (10) | 5 | Multiple ligand Emax model | 171.012 | 4.525 |
*
Lowest value for likelihood (indicating best fit) is shown in italics and the most parsimonious model (the best fit amongst nonstatistically different lowest-LL which has fewer number of model parameters) is shown in bold italics.
**
−LL = – log likelihood; Model 3 produced statistically significant better fit than all models except models 1, 2, 4 and 9. The differences between −LL of the latter four models and model 3 was not significant.