I have collected data at 3 different time periods. For the 1st time period, I have analysed using SEM (SPSS + AMOS). Subsequently, I have collected data for 3 different time periods (say each year). What type of statistical analysis can be done for drawing inference ?
Your question hides more than it reveals. You need to mention along with disclosing your constructs whether all of them were measured at three different times or IVS were measured at time 1, mediators at time 2 and DVs at time 3. Additionally, your sharing the hypotheses of the study would enable the community to share their expertise to enable you to successfully go with your analysis.
The type of analysis would depend on what research questions you would like to address. A common situation is when you have a number of predictor variables at baseline (or do not vary across time, such as a persons sex) and you want to see how they predict how change in some outcome variable which is subsequently measured at a number of measurement occasions (waves). Although not the only acceptable approach, a very common method is to use linear mixed modelling, LMM (or generalised LMM for non-normally distributed outcome variables) with a random intercept and slope. The main advantage of this method over older, or "traditional" methods which only use cases with complete data (eg repeated analysis of variance) is that it is very common that there is non-random attrition over time, and that can lead to bias in the results. Although the use of (G)LMMs does not necessarily eliminate completely attrition bias except under vary special circumstances, using that procedure is generally regarded as a reasons step in reducing the bias due to attrition bias.
Other procedures which are also commonly used and can address the problem of attrition bias, are structural equation modelling (SEM) with full-information maximum likelihood estimation FIML), or the use of multiple imputation. There are more advanced procedures that can, theoretically, better address the problem of attrition bias (such as joint modelling) but they are not always successful, and usually require large samples.
@Raja In addition to what @John has suggested, you can also look for Multi-Group SEM or Cross Lagged Panel Analysis (In case the nature of your data is of repeated measures).
Why you needed to collect data at three different time periods? Your answer lies in there!
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS