I am currently planning out my masters' thesis design and will be working with a large data set with variables that have been meseasured at two time points. Intially I thought of just using a general cross lagged panel model, however after some reading I understood that a RI-CLPM controls for stable trait variables. Though it seems that the model is identified with at least three time points? I cannot find a direct clear answer to this question.
Emily Gabriela Vira if the time intervals between subsequent occasions vary, then you need to estimate different autoregressive and cross-lagged effects between each pair of adjacent measurements, constraining the lagged regression coefficients leads to an uninterpretable blend of different lagged relationships. Based on the above scenario. Is the lagged parameters are invariant over time? This will typically not be true for the standardized lagged parameters, because these are a function of the within-unit variance of the predictor and the within-unit variance of the outcome. As these variances are typically not (constrained to be) equal across the occasions (which is complicated due to the recursiveness in the model), the standardized lagged parameters can differ even if the unstandardized lagged parameters are constrained to be the same.
What about measurement error? Failing to account for measurement error can bias lagged-parameter estimates downward, leading to a loss of power You also need to look at a model for configural invariance. On the other hand, non-significant result implies that the current model does not have to be rejected, as there is measurement invariance across structure and fluctuating within-structure. Is the chi-square test significant? The procedure for establishing measurement invariance relies heavily on chi-square difference testing which, as mentioned before, can have serious disadvantages such as an increased Type I and Type II error rate when the base model is misspecified.
Hope this helps. Good luck.
Yes, you would need at least three time points to fit the lagged model. With only two time points, it is more of pre and post analysis. You can implement random intercept and slope model by re-arranging your data from wide to long format and run linear mixed effect model. If you are familiar with R, use NLME and LMMEANS packages (Bob Winter wrote few easy tutorials). You can also refer to: https://m-clark.github.io/docs/mixedModels/anovamixed.html for a good tutorial. If you use SAS, search for 'linear mixed effects' at Lexjansen page). If you use SPSS, try IDRE UCLA website. Good luck.
Emily Gabriela Vira if the time intervals between subsequent occasions vary, then you need to estimate different autoregressive and cross-lagged effects between each pair of adjacent measurements, constraining the lagged regression coefficients leads to an uninterpretable blend of different lagged relationships. Based on the above scenario. Is the lagged parameters are invariant over time? This will typically not be true for the standardized lagged parameters, because these are a function of the within-unit variance of the predictor and the within-unit variance of the outcome. As these variances are typically not (constrained to be) equal across the occasions (which is complicated due to the recursiveness in the model), the standardized lagged parameters can differ even if the unstandardized lagged parameters are constrained to be the same.
What about measurement error? Failing to account for measurement error can bias lagged-parameter estimates downward, leading to a loss of power You also need to look at a model for configural invariance. On the other hand, non-significant result implies that the current model does not have to be rejected, as there is measurement invariance across structure and fluctuating within-structure. Is the chi-square test significant? The procedure for establishing measurement invariance relies heavily on chi-square difference testing which, as mentioned before, can have serious disadvantages such as an increased Type I and Type II error rate when the base model is misspecified.
Hope this helps. Good luck.
Pradeep Paraman Thank you for answer as well! I do not have access to the full data set yet, but I will be checking for measurement invariance. Configural will be the minimum to achieve, I will then assess metric and scalar invariance, and to avoid the biases of chi-square, a change of -.01 in CFI will be used.
- Badges
- Science topic
- Similar topics
- Thesis Research