Hello,
I am currently analyzing physical performance trajectories using a group-based semi-parametric trajectory model. Model adequacy was assessed according to the diagnostic criteria proposed by Nagin (2005).
The selected trajectory model met three of the four recommended diagnostics:
(1) Average Posterior Probability of Assignment (AvePP ≥ 0.70 for all groups);
(3) Close correspondence between estimated group membership probabilities and the observed proportions assigned to each group; and
(4) Relatively narrow confidence intervals for group membership probabilities.
However, for the "intermediate" trajectory did not reach the commonly recommended threshold of 5.0 for the Odds of Correct Classification (OCC), while all other diagnostics indicated adequate model performance.
I would appreciate any insights on potential reasons for this result. In particular, could a lower OCC for an intermediate trajectory reflect partial overlap or gradual transitions between adjacent trajectories in continuous outcomes such as physical performance? Are there other methodological or substantive explanations that could account for an OCC below 5 despite satisfactory performance on the remaining Nagin diagnostics?
Thank you very much for your time and insights.
Best regards,
Laíza
I am currently analyzing physical performance trajectories using a group-based semi-parametric trajectory model. Model adequacy was assessed according to the diagnostic criteria proposed by Nagin (2005).
The selected trajectory model met three of the four recommended diagnostics:
(1) Average Posterior Probability of Assignment (AvePP ≥ 0.70 for all groups);
(3) Close correspondence between estimated group membership probabilities and the observed proportions assigned to each group; and
(4) Relatively narrow confidence intervals for group membership probabilities.
However, for the "intermediate" trajectory did not reach the commonly recommended threshold of 5.0 for the Odds of Correct Classification (OCC), while all other diagnostics indicated adequate model performance.
I would appreciate any insights on potential reasons for this result. In particular, could a lower OCC for an intermediate trajectory reflect partial overlap or gradual transitions between adjacent trajectories in continuous outcomes such as physical performance? Are there other methodological or substantive explanations that could account for an OCC below 5 despite satisfactory performance on the remaining Nagin diagnostics?
Thank you very much for your time and insights.
Best regards,
Laíza
