Health and socio-economic outcome data from the better quality studies (n = 23/34) were mapped onto a one page logic models by two independent reviewers and a final model reflecting reviewer agreement was prepared.
DAMMIN then produced a final model by minimising differences between experimental data and scattering of the model.
Furthermore, in order to retain a final model, a two-by-two interaction terms were checked for variables that remained statistically significant in the pre-final model.
We also computed the overall estimate by combining NHANES cohort data in a final model additionally adjusted for cohort; the predicted probabilities for these models were computed for a prototype participant as defined above.
These scales were analysed with structural equations modelling (LISREL), by modifying a start model into a final model with a good fit, within theoretical constraints.
All variables selected by this algorithm were included in a final model in order to get odds ratio estimates with the corresponding 95%% confidence intervals (Table 4).
Using this approach, several model specifications are first estimated (there are as many models as possible combinations between the explanatory variables) and a final model computed by using the average-weighted parameters and SEs obtained from the different model specifications.
A final model was achieved by entering the variables retained in the backward selection model.
Fig. 7 (a) Final model of SPr2 obtained by tomographic inversion.
A final model was then developed by backward selection of factors that showed p<0.05.
