Level 1 Model Statistics for BSI GSI Global Scores to Determine Best Fitting Change Function
| Model Estimated | −2LLR | AIC |
|---|---|---|
| Cubic polynomial: Int, T1, T2, T3 random* | 387.87 | 417.87 |
| Cubic polynomial: Int, T1, T2 random | 387.70 | 411.70 |
| Cubic polynomial: Int, T1 random | 407.92 | 425.92 |
| Cubic polynomial: Int random | 442.20 | 456.20 |
| Quadratic polynomial: Int, T1, T2 random | 391.71 | 413.71 |
| Quadratic polynomial: Int, T1 random | 411.52 | 427.52 |
| Quadratic polynomial: Int random | 445.38 | 457.38 |
| Logarithmic Model: Int, TLN random | 387.68 | 401.68 |
| Logarithmic Model: Int random | 488.72 | 498.72 |
| Linear: Int, T1 random | 425.94 | 439.94 |
| Linear: Int random | 458.39 | 468.39 |
| Intercept-only model (no change) | 521.46 | 529.46 |
−2LLR, log likelihood ratio statistic; AIC, Akaike's information criterion: smaller values are better. Int, Intercept coded so it is an estimate of BSI GSI at first assessment period. T1, Time coded for linear function. T2, Time squared coded for quadratic function. T3, Time cubed coded for cubic function. TLN, Time coded for logarithmic function. Time is number of months since initial assessment.
↵* Hessian matrix was positive definite although all convergence criteria were met.