Case: Longitudinal research on kids over a period two years from 10 to 12 standard to analyse their behaviour toward education (sample size around 1200 during 10th and for 12th standard reduced to 1150.
Condition:
Approach 1:
Approach 2:
Kindly guide me how to proceed in such condition when data is longitudinal in nature.
Manvendra,
could you describe what you mean by "validate"? If you need to, e.g., validate your measurement instruments, then I would suggest doing this in the same sample you use for inference would be inappropriate.
In terms of what model to use for Step 2, this paper might help:
Liu, Y., Mo, S., Song, Y., & Wang, M. (2016). Longitudinal Analysis in Occupational Health Psychology: A Review and Tutorial of Three Longitudinal Modeling Techniques. Applied Psychology An International Review, 65(2), 379-411.
@Heiko Breitsohl Validation means to test the best-fitted model for each case and to validate the framework.
2. to find the difference between their behaviour using ANOVA
If you have multiple indicators of the construct in question (i.e. with repeated measurements; 10 vs. 12) you may employ a so-called latent difference score model in AMOS. McArdle, J. J., & Nesselroade, J. R. (1994). Using multivariate data to structure developmental change. In S. H. Cohen & H. W. Reese (Eds.), Life span developmental psychology: Methodological contributions (pp. 223-267). Hillsdale, NJ: Erlbaum.
Hello Manvendra. I think what you are asking is do you need the models to 'fit' for both 10th and 12th standard timepoints. The answer would be yes; if the model fits the data for only one timepoint then it is not robust (sample specific) or sensitive to changed understanding of the constructs by the participants. You probably would not expect that, if we are talking about adolescents/teenagers over 2 years.
Picking up on Michael's comment, if this involves multiple indicators, then you will also need to test for measurement invariance across time (along with the standard factor reliability for each factor at each timepoint). If this is a Path analysis (i.e, single observed variables), then you are just considering model fit statistics.
Wishing you well.
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Structural Equation Modeling